Datasets:

Vereye
/

benchmark-synthid-image-protocol

set string	photo int64	transformation string	mode string	group string	marked int64	label string	params string	pixel_sha256 string	file_sha256 string	verdict string	identified int64	owners_named int64	crc_keys int64	ms int64	file float64
new-coco-val2017-512	21,839	brightness	random	color	1	factor1.157	{"brightness":1.1568272767003904}	673d38c4b3989de4ff896004d09eb867894fd0dd7764f422e190b20e2c876152	cc1818a2c5ef3812cae31b0a0620df36a7bfa16615002869a23a7bcf7af82169	MARK FOUND	1	1	1	935	null
new-coco-val2017-512	21,839	brightness	random	color	0	factor1.157	{"brightness":1.1568272767003904}	0fbb721725af2995ec05b4c2cbb84a9b6a7efdc1815bc7e8699679eb9d6c364f	e43817787ce66e7981bb9336bac07ae222703d5e19e1361ed06a0f3ab2985d20	NO MARK FOUND	0	0	0	674	null
new-coco-val2017-512	21,839	brightness	worst	color	1	factor0.5	{"brightness":0.5}	7aa7631c5a6561e8644c6e92420ad6e59ae7c9cc0816db95351c28544d05e1b0	64af72720d6400e0ab91c663c8bb184bfeb0c663d1a157398f9bba340e6e0e26	MARK FOUND	1	1	1	954	null
new-coco-val2017-512	21,839	brightness	worst	color	0	factor0.5	{"brightness":0.5}	19146f0dd873fccffb89938cc81d07fb18f2da0bce81158ab7bb706c7a1c901b	b8acb6cb192656b0a7bfeae991b24f898ccb7cda645f81de3fd50ab225ac369a	NO MARK FOUND	0	0	0	631	null
new-coco-val2017-512	21,839	contrast	random	color	1	factor1.249	{"contrast":1.2487891880098259}	fa577b9b4bc5359c861bfeb1c93a128cb9b33ca6724d9490c0fbedaf66d26291	dc8312ee626c42f4faf8418515fa97042c21d27d08aa202f19250eb7271a588c	MARK FOUND	1	1	1	491	null
new-coco-val2017-512	21,839	contrast	random	color	0	factor1.249	{"contrast":1.2487891880098259}	c18d3deb80e6a9868421a1c78bb5b0b604e11cfb05d82b2c910d4dcfeecbb30c	418a1b69a76b857fa68616793116492405087b7fea78634108872088d7a71ff9	NO MARK FOUND	0	0	0	677	null
new-coco-val2017-512	21,839	contrast	worst	color	1	factor0.5	{"contrast":0.5}	9bea385351e186a0763af7f99fb3817d21c958143550f609b3b9179021a993d1	a5fb2aacc5879b3a35ea4f9a80a1b2c555dc9f421950902bd9db3148eb869734	MARK FOUND	1	1	1	477	null
new-coco-val2017-512	21,839	contrast	worst	color	0	factor0.5	{"contrast":0.5}	972836951c1d844cdc5cc61e67a304e4d468a77e337b63cc92aa70f76bb12617	21ae5da8fbe3406d190639638aa11a1b1f4da7583e4ecadca99b9057e5af1ec5	NO MARK FOUND	0	0	0	643	null
new-coco-val2017-512	21,839	exposure	random	color	1	ev-0.197	{"exposure_ev":-0.19652437957427393}	c468cebff383039b12e733c9d9fa8612d258bff5d62964042e36f0fe97cce52b	437540bb7fce20fe32996c9dc48800ec228ea5ff972e8da23c575b58dc427f33	MARK FOUND	1	1	1	480	null
new-coco-val2017-512	21,839	exposure	random	color	0	ev-0.197	{"exposure_ev":-0.19652437957427393}	76d4fac7982145a7754edf1275cf88bc7d8302d548fd149177df71ffba540c4e	1c63a2456e09a5424d25d814d82efcb11b319c563914585b8baf8392b9604fdc	NO MARK FOUND	0	0	0	588	null
new-coco-val2017-512	21,839	exposure	worst	color	1	ev1	{"exposure_ev":1}	5f8d7d28292c6e966a2630709e7c921f0dfcf8ec5d1638eec55eb5e197ea8319	8c29178d09943b9227d3344862609c66dcfc2726e86435aba0981a4cbe7564a2	MARK FOUND	1	1	1	480	null
new-coco-val2017-512	21,839	exposure	worst	color	0	ev1	{"exposure_ev":1}	835f31565118f9b8bbe0b877509587921bf1aa9a425f7cf26ea3ce78d4bdade8	cd63145581b260ce20235bda31a15b95489a038a2a4a92cb746f2ac16d43bcc9	NO MARK FOUND	0	0	0	618	null
new-coco-val2017-512	21,839	grayscale	random	color	1	luma601	{"grayscale":"ITU-R 601-2 luma (PIL ImageOps.grayscale)"}	db2b47415787e55219d54ff464baa1faccd972517c133312f3803bf6ee4f1761	72112a24268d6934edbbc923eba41290baf81ce9b6f48b2c6233963fac4038da	MARK FOUND	1	1	1	1,532	null
new-coco-val2017-512	21,839	grayscale	random	color	0	luma601	{"grayscale":"ITU-R 601-2 luma (PIL ImageOps.grayscale)"}	e7da55f2c0964f021f2d5cfae8cf59389a57308553bac54c336996fe5e1fc24f	c7d31cd40198798e07cb9cd672a2859d4b7a9010b55b151e203ee6e081486ca9	NO MARK FOUND	0	0	0	1,082	null
new-coco-val2017-512	21,839	grayscale	worst	color	1	luma601	{"grayscale":"ITU-R 601-2 luma (PIL ImageOps.grayscale)"}	db2b47415787e55219d54ff464baa1faccd972517c133312f3803bf6ee4f1761	72112a24268d6934edbbc923eba41290baf81ce9b6f48b2c6233963fac4038da	MARK FOUND	1	1	1	1,383	null
new-coco-val2017-512	21,839	grayscale	worst	color	0	luma601	{"grayscale":"ITU-R 601-2 luma (PIL ImageOps.grayscale)"}	e7da55f2c0964f021f2d5cfae8cf59389a57308553bac54c336996fe5e1fc24f	c7d31cd40198798e07cb9cd672a2859d4b7a9010b55b151e203ee6e081486ca9	NO MARK FOUND	0	0	0	1,091	null
new-coco-val2017-512	21,839	hue	random	color	1	shift0.085	{"hue_shift":0.08503136957606333}	ad76f0a15e9a70e9b38c75463171c14afd92abe0a3630594c515d5d671f8bb3e	5dfd7c54c18320c760ab8dbc13496a1a75153f7d0d80722134d0cdb3ee270fa4	MARK FOUND	1	1	1	508	null
new-coco-val2017-512	21,839	hue	random	color	0	shift0.085	{"hue_shift":0.08503136957606333}	68bc1e13e4ff7990a043de4730254cfa387eab017d79e852fe79948c6b23092c	572b58ff11e2975fc5ffbbba7e00db572803f4ac84875973b4f0cbfae4bc0732	NO MARK FOUND	0	0	0	647	null
new-coco-val2017-512	21,839	hue	worst	color	1	shift0.1	{"hue_shift":0.1}	39230623eb3f9935294668bc97995df93e78a1ad5fda55f9288d3005e4d893df	e5501eaeb2b288cdbcd948151ced388699a043eef0b2fb51fd5c8874dea03bab	MARK FOUND	1	1	1	510	null
new-coco-val2017-512	21,839	hue	worst	color	0	shift0.1	{"hue_shift":0.1}	21bcc00e75b965eb6394c5218586e542beed4fd7794dc2f38c0d24b5c3e51569	37ae19f64eef3453dc8577e5f45cbccd916789971b04c1ab92c9d32ca3bcb474	NO MARK FOUND	0	0	0	659	null
new-coco-val2017-512	21,839	saturation	random	color	1	factor0.781	{"saturation":0.7809761440112882}	48ae58c3d2c67baf942da347234878966364cd4e5ec81d0786a5c1df5e05ee17	2ed3f6f093433865e70c9602693541a9778b3edadeea01f767fbbbf331924d10	MARK FOUND	1	1	1	449	null
new-coco-val2017-512	21,839	saturation	random	color	0	factor0.781	{"saturation":0.7809761440112882}	35c1e621f474f74f58b82002f3ebd882f5c8434c47f4773b5ed88ebf898b9b21	3fab8786590287543ca722ecee6dc54c7daa2bea6c688d07bd62da22c949079c	NO MARK FOUND	0	0	0	663	null
new-coco-val2017-512	21,839	saturation	worst	color	1	factor0.5	{"saturation":0.5}	adda5379c2fcb0a16a72bbbc4b42b06179ede55051430cb2d6e6ba2d83ad6234	5d4213c872f2e5df660b44851bc656c950ee80f304d4736a5f268ba76230e258	MARK FOUND	1	1	1	260	null
new-coco-val2017-512	21,839	saturation	worst	color	0	factor0.5	{"saturation":0.5}	f168a506cc4415b972d4ce1adbee7abb854733ae54965a807710d4c491c841e4	2d48435ea7253d35e4fb4e07a59c7ec38f9adea7a2308b8acb73440d41d5639d	NO MARK FOUND	0	0	0	671	null
new-coco-val2017-512	21,839	sharpness	random	color	1	factor2.197	{"sharpness":2.1973010884936395}	0dba0fd95f9de3971d7e2d07d226e67831219bc31c4a3308ef19f1ac6872d6a8	a401299944764bc83802f71fffc84589e2474da6aa6ed84b079fa0c667325cb4	MARK FOUND	1	1	1	533	null
new-coco-val2017-512	21,839	sharpness	random	color	0	factor2.197	{"sharpness":2.1973010884936395}	35165d4cfa006e422c0a7c43e4f1a46fc159ecade349d0bdd0ec295fe8619a61	c25a48fcceda674864799bddf5fb2b9128fc687e5ba490044fb8232c17859373	NO MARK FOUND	0	0	0	658	null
new-coco-val2017-512	21,839	sharpness	worst	color	1	factor0	{"sharpness":0}	a5500e544913e036d3df4daf91bf930a349df8cc30125154dbe1db2c7b1e5b60	d25abe6c3318e8844a4a8b829a8d104dd5f901b0e0e044e6f2123df52df7f971	MARK FOUND	1	1	1	545	null
new-coco-val2017-512	21,839	sharpness	worst	color	0	factor0	{"sharpness":0}	29e21a9ee67342f4974a775266001d86ea3de28ed6e150b1780966399e460537	8beeb196867b53dfec49d53a2e965adf981117f3dbddbbfe7b18ffd3cc40a30e	NO MARK FOUND	0	0	0	672	null
new-coco-val2017-512	21,839	combined	random	combination	1	crop0.777-bright0.882-contrast0.926-jpegq57	{"crop_area":0.7771023467573615,"crop_box":[53,23,504,474],"brightness":0.8818771600811913,"contrast":0.9264633536401312,"jpeg_quality":57}	2f13781cdf99cd73327efd9eb55175e9128e26241b75a514e3b16d972b99a1d9	5616e83d014f9f83eb8ddcbf5e03aef370481e4922b443e09f8387562aa4825e	MARK FOUND	1	1	1	433	null
new-coco-val2017-512	21,839	combined	random	combination	0	crop0.777-bright0.882-contrast0.926-jpegq57	{"crop_area":0.7771023467573615,"crop_box":[53,23,504,474],"brightness":0.8818771600811913,"contrast":0.9264633536401312,"jpeg_quality":57}	d5401461d18a96c3cd3706d9458d16f3fb867bb6c7aea6fa6ebb2fd76097e9b6	efdf59e87a81a16c795f2b3eeb475788b449c2a9e8c622716b0590bee0e9ada9	NO MARK FOUND	0	0	0	624	null
new-coco-val2017-512	21,839	combined	worst	combination	1	crop0.5-bright0.7-contrast0.7-jpegq50	{"crop_area":0.5,"crop_box":[141,131,503,493],"brightness":0.7,"contrast":0.7,"jpeg_quality":50}	ad024dfb82fdeb96e682da74a9a46aa900ba7c766415d1a75eff695b4cd18d95	b1efc4e8108facab5a9f231dfa3794265e0ceae30aa5879e03bf061703d09e20	MARK FOUND	1	1	1	871	null
new-coco-val2017-512	21,839	combined	worst	combination	0	crop0.5-bright0.7-contrast0.7-jpegq50	{"crop_area":0.5,"crop_box":[141,131,503,493],"brightness":0.7,"contrast":0.7,"jpeg_quality":50}	c8fe377dc0dfd79e9619a815d1c544022320be84564565d38a496ced494f16f9	8e0785ca629c3f144aac4b32e9bbb9f2261df3a2537522ec665a40fbd684c48d	NO MARK FOUND	0	0	0	707	null
new-coco-val2017-512	21,839	combined_norotate_nocrop	random	combination	1	bright0.957-contrast0.732-jpegq53	{"brightness":0.9571890612755605,"contrast":0.7318615810542541,"jpeg_quality":53}	c97b0cd7d9fd2817e2087f020af9339571c5e66aa8425c627ee27e9b11984aaf	68175416397da233c4a3aaf360b166de4ac54da07e3ef68d1d716cd38dcc25b9	MARK FOUND	1	1	1	492	null
new-coco-val2017-512	21,839	combined_norotate_nocrop	random	combination	0	bright0.957-contrast0.732-jpegq53	{"brightness":0.9571890612755605,"contrast":0.7318615810542541,"jpeg_quality":53}	c6e853541645e6381a916d0503aa3ca77033513af087976b50f50997888272fe	72e1f2fd64bd04d279926ffa09f9fa535e3d3a84c01f849fc87237a687a44c29	NO MARK FOUND	0	0	0	617	null
new-coco-val2017-512	21,839	combined_norotate_nocrop	worst	combination	1	bright0.7-contrast0.7-jpegq50	{"brightness":0.7,"contrast":0.7,"jpeg_quality":50}	8896f0eb092b211635a82998bc9f2af7995f3b57700fdea26bb7819b7fa9426f	16e72ec2f1fb24135f0503eaf13c46bc66ddf12485308f57b53b6c082fb25b30	MARK FOUND	1	1	1	388	null
new-coco-val2017-512	21,839	combined_norotate_nocrop	worst	combination	0	bright0.7-contrast0.7-jpegq50	{"brightness":0.7,"contrast":0.7,"jpeg_quality":50}	60906e2a575f026bdb99c342959f91bf669e31d7dca07b6004dcc8e6c9c34cc1	50b37128cb48cd47494bb2fa469427280ef90dd464f01defb19d854eb31a82ae	NO MARK FOUND	0	0	0	640	null
new-coco-val2017-512	21,839	combined_nocrop_rotate	random	combination	1	rot9.88deg-bright0.794-contrast1.235-jpegq82	{"rotate_deg":9.876763655079685,"brightness":0.7942886510630675,"contrast":1.2353503312617893,"jpeg_quality":82}	d4dd7cfba6005e419dbe2fbb4bac7572de3c5380e59fafb33f69081c12dcd317	746af2bccdf7ff0f9527b093c9df79474cb2eba13e87e0de17223380ca39c1f1	MARK FOUND	1	1	1	568	null
new-coco-val2017-512	21,839	combined_nocrop_rotate	random	combination	0	rot9.88deg-bright0.794-contrast1.235-jpegq82	{"rotate_deg":9.876763655079685,"brightness":0.7942886510630675,"contrast":1.2353503312617893,"jpeg_quality":82}	4ae02b5c3833d5a76fe1a6c7993be4c8873f8c7064f0a803580bb94d7b7efbb0	05faa15cb2f4e6e9e1c08d93855d9561c71cbefab3f45e54adc53b9e523c3912	NO MARK FOUND	0	0	0	636	null
new-coco-val2017-512	21,839	combined_nocrop_rotate	worst	combination	1	rot10deg-bright0.7-contrast0.7-jpegq50	{"rotate_deg":10,"brightness":0.7,"contrast":0.7,"jpeg_quality":50}	b9454dec32c51795dd213a1f390bf557fbe28d6e036cb6dceb019ab355ee4a5c	bfe9dbc1d41dc6b305c1b49425328b47b51d45b880034d40f26580ad18910ac7	MARK FOUND	1	1	1	1,034	null
new-coco-val2017-512	21,839	combined_nocrop_rotate	worst	combination	0	rot10deg-bright0.7-contrast0.7-jpegq50	{"rotate_deg":10,"brightness":0.7,"contrast":0.7,"jpeg_quality":50}	1532fd54ab658ef45c56deed7e7db4535689dbd4bd1b4d2fe6a128e93a2510b8	21b38664b5cbd6e874faa9358b0fc9bbe4e487d93bef2d46ac62120d01c45556	NO MARK FOUND	0	0	0	625	null
new-coco-val2017-512	21,839	combined_rotate	random	combination	1	crop0.735-rot1.83deg-bright0.732-contrast1.018-jpegq55	{"crop_area":0.7347804730462671,"crop_box":[60,70,498,508],"rotate_deg":1.8319485375774285,"brightness":0.7323807259206081,"contrast":1.0182722782760927,"jpeg_quality":55}	21ae43a777e1de014850b2154cb35461837d7b285f50dae56fb01928ac44035a	d1aded00fb42b5765376e8da1ea05b4efdb13ab65ea6611a0e6bd73a135225a9	MARK FOUND	1	1	1	487	null
new-coco-val2017-512	21,839	combined_rotate	random	combination	0	crop0.735-rot1.83deg-bright0.732-contrast1.018-jpegq55	{"crop_area":0.7347804730462671,"crop_box":[60,70,498,508],"rotate_deg":1.8319485375774285,"brightness":0.7323807259206081,"contrast":1.0182722782760927,"jpeg_quality":55}	a4c55fb13343c890b3cc70cf7d139cbce79969fc7cefb4e1fad43e1dde6d74ad	7ad4c13bcd3d6ba444212ad3e82aa09ccde5e28f4ea7f3a097f1b646f71f041d	NO MARK FOUND	0	0	0	662	null
new-coco-val2017-512	21,839	combined_rotate	worst	combination	1	crop0.5-rot10deg-bright0.7-contrast0.7-jpegq50	{"crop_area":0.5,"crop_box":[80,143,442,505],"rotate_deg":10,"brightness":0.7,"contrast":0.7,"jpeg_quality":50}	6744846193813f2638610d69e752c168e98c3002561b641b727e4c6a8a8a3210	14c2862a3d20060486e2f531bda3fca63416eb9515854dbf3ee77910dcbe9fd2	MARK FOUND	1	1	1	463	null
new-coco-val2017-512	21,839	combined_rotate	worst	combination	0	crop0.5-rot10deg-bright0.7-contrast0.7-jpegq50	{"crop_area":0.5,"crop_box":[80,143,442,505],"rotate_deg":10,"brightness":0.7,"contrast":0.7,"jpeg_quality":50}	8f17cef59f9a30acbf7b467fc9405a22458f741bb7dfec344f09e1ac81e3d782	90024bc87116957fe81dbe6d42e73f2a78ceca39d509f9e9d40b7027b2bc73a8	NO MARK FOUND	0	0	0	637	null
new-coco-val2017-512	21,839	gaussian_noise	random	noise	1	sd17.96	{"gaussian_sd":17.959584344389494}	cb91229a2a770e527567be6e99874c9f8415398f644bb2fa13c00639352f8e19	dd685f518ac2da4b061a9572e4bedf0981290a0120a9f215c06ff101ffc5caa3	MARK FOUND	1	1	1	500	null
new-coco-val2017-512	21,839	gaussian_noise	random	noise	0	sd17.96	{"gaussian_sd":17.959584344389494}	b6fb0b5e6dcc1ca1cc6485a93514c23e2166a7260c54c683118b49d63e696ac1	c8dc19094495454694aba1e21f449397f3ef89cd30d85def22b8911e2e27c061	NO MARK FOUND	0	0	0	642	null
new-coco-val2017-512	21,839	gaussian_noise	worst	noise	1	sd20	{"gaussian_sd":20}	c7fce630f116708ed2613ddb32ffaa1e5ea314a05848d1d40c0be634e656f3d8	2922d512028baeb0fcd97d6cb881f11bc080c61bf667318e0d0fc6bbbd670f83	MARK FOUND	1	1	1	486	null
new-coco-val2017-512	21,839	gaussian_noise	worst	noise	0	sd20	{"gaussian_sd":20}	4afaf99845f75f4e22e514cdbb5d7fd0167993d814ff3b131e0e01514696d32c	567337e5f23911120b9be10f62438552850896a8a938352ac852d0ffb5dd3a17	NO MARK FOUND	0	0	0	640	null
new-coco-val2017-512	21,839	impulse_noise	random	noise	1	p0.0355	{"impulse_p":0.03552064025634522}	114de4e534c52489f7958d14be9951f32e3c39ce5bc4b338383c9e83f79e4720	2ad68861ca483a31ad8fdc0f09961fe64349172513a83ace39fb5143aaeb9681	MARK FOUND	1	1	1	243	null
new-coco-val2017-512	21,839	impulse_noise	random	noise	0	p0.0355	{"impulse_p":0.03552064025634522}	848dd78a6b4cdbdafbc036ee8224418594d7f2214ab9238ad7869ca70c5dece1	728ae88c2b8d237dcb24066cda5aa99367f5d78198d8bc118334821c0ebca712	NO MARK FOUND	0	0	0	606	null
new-coco-val2017-512	21,839	impulse_noise	worst	noise	1	p0.05	{"impulse_p":0.05}	f7ca40f00903da87f0d5cc25dadabac146abc5ed58d0662caa0ceff0b54b1144	98ee1ca8c82307a20bf24505364ec88a47dc250c7ffdfcfd44a0b9fa1e1eec2a	MARK FOUND	1	1	1	222	null
new-coco-val2017-512	21,839	impulse_noise	worst	noise	0	p0.05	{"impulse_p":0.05}	2853b5bc14dabc46e54f1e385d1cc4bf3c598ec93572e42a779018c1bd3a8274	92fc25f21a04cd122c8d1bf605e3d47c354ba42d95541ee2d4d99b36d6a2ff3a	NO MARK FOUND	0	0	0	610	null
new-coco-val2017-512	21,839	shot_noise	random	noise	1	k318.4	{"shot_k":318.36693635025534}	ca6e2bc9614580be84a0ca685f1b1fdd63c577ea5b28c628f28be9938c170846	cb8253014d820b02e9c84973e330921d53f3c0c9c635109581608682d4892357	MARK FOUND	1	1	1	468	null
new-coco-val2017-512	21,839	shot_noise	random	noise	0	k318.4	{"shot_k":318.36693635025534}	270ea90576ad808ac29efcb8fad6a42000f7eb1d501c8e2d36dd3a1d2d16ba21	0a4aa00fd8c54abeb4ad78b4bd73d254a5ad9093f7095c529cdfddfa4ce32ee1	NO MARK FOUND	0	0	0	640	null
new-coco-val2017-512	21,839	shot_noise	worst	noise	1	k60	{"shot_k":60}	8d13414aaee25b053edd6b684e8fbe1b34e304b2120ba6ae50d82183b9e11def	ef9133b442e426fb9baa74b4b7362df7f90e1147e4904dc04aa7654d1f43b769	MARK FOUND	1	1	1	489	null
new-coco-val2017-512	21,839	shot_noise	worst	noise	0	k60	{"shot_k":60}	5661af9ff58aab7722d026f996374beb6aaa00b3c08d538e6a458d2496ba0edd	60cb05767090e6eed91865ebb0030ad7f7f847f8a5fa8fa779b92c3fa1fa91ae	NO MARK FOUND	0	0	0	628	null
new-coco-val2017-512	21,839	speckle_noise	random	noise	1	sd0.069	{"speckle_sd":0.06944142053043521}	317c411bc48ae27d154192495612eb38b8297454b1bcd6dd555894d6c1113912	b8ff5de52949cd74fdac2e77ec3fc0877dfbf37f6ca043aa8b591e13ca73a333	MARK FOUND	1	1	1	521	null
new-coco-val2017-512	21,839	speckle_noise	random	noise	0	sd0.069	{"speckle_sd":0.06944142053043521}	dea8d83a4529aba81688a02b48a7794139a53708e1ed6bc95b2e2799f73518b6	6cbb60398858623d3ab8c3666a22421ae677a752c6833bfab413c6360e98dd64	NO MARK FOUND	0	0	0	548	null
new-coco-val2017-512	21,839	speckle_noise	worst	noise	1	sd0.2	{"speckle_sd":0.2}	1d3e688d1d3e87ef4690b0543c2ee8d90e1d48ed365c6069433da5a6a854339f	9349fb49a8ef421248697585d0014cbf6def860c7e3cecfdb07aabb77662de3b	MARK FOUND	1	1	1	639	null
new-coco-val2017-512	21,839	speckle_noise	worst	noise	0	sd0.2	{"speckle_sd":0.2}	bb38387ea56978757e444f598a7887d3ccb05491e2b96fbfa8f766c5e847d71d	631695256cd61a0dbb9d03654a49301a46333e79456075c2e308d0679750faee	NO MARK FOUND	0	0	0	743	null
new-coco-val2017-512	21,839	emoji_overlay	random	overlay	1	area0.052	{"emoji_area":0.051796337859320196,"emoji_box":[365,209,481,325]}	b15328ccd2ec1786c56bf351f5c4d420276d94fb1c5f247a04bce89d20b5275b	f6dbaadde205f7ee3ab8fb88419a89611a6df0f2d75edcd700389e1f5ddeba41	MARK FOUND	1	1	1	504	null
new-coco-val2017-512	21,839	emoji_overlay	random	overlay	0	area0.052	{"emoji_area":0.051796337859320196,"emoji_box":[365,209,481,325]}	fed816418d4e8b3b5855ff41b826631aa66f6246f837f1b274d25a7e37398416	d791e20e0a9f8aaed922f6070dd3547fc3489d00b14a71c45013d81b7a8aa4e8	NO MARK FOUND	0	0	0	568	null
new-coco-val2017-512	21,839	emoji_overlay	worst	overlay	1	area0.15	{"emoji_area":0.15,"emoji_box":[94,55,292,253]}	47fe5a1458adf5619f711ba26903dc13db943a512bdb0beed2b7ffe5d075c5a9	925d7cf5209f7436ce39fd83f4fc0a22906fc15a43669c17d59365f5c5b784f1	MARK FOUND	1	1	1	323	null
new-coco-val2017-512	21,839	emoji_overlay	worst	overlay	0	area0.15	{"emoji_area":0.15,"emoji_box":[94,55,292,253]}	699394ba3cc441c24d33fc5e7df3eb4ee3757ca3d0cb32dc2c95c8688057d1cd	1c53f31c18a065181592339145dfabd764b48686169660313ab2dabd50b73b6e	NO MARK FOUND	0	0	0	587	null
new-coco-val2017-512	21,839	instagram	random	overlay	1	lark	{"filter":"lark"}	5c7a1558d055e2311a3ae2e42358cc14aca0c47368ced13741470230e246b3f0	9c8adb4a335bc4cafcab64a89a69096f653ac8b0c54f2baf85a6376cdaeec2df	MARK FOUND	1	1	1	496	null
new-coco-val2017-512	21,839	instagram	random	overlay	0	lark	{"filter":"lark"}	50752bb81b799c01a489a4f44520de57fe71cbe7a109f266e6a74e5b8198515d	6850c27c23173f51c386fbd6737b573c375ebfd57443b5587d69fdeede273cd2	NO MARK FOUND	0	0	0	678	null
new-coco-val2017-512	21,839	instagram	worst	overlay	1	reyes	{"filter":"reyes"}	c5026e4590cdd74015cc10e873c9cf6f7235e71f1fff50b04530108d56b630c1	71671c756524df9f10ffac370d8ec7fa43a7879801be2c3326d51119c47b50c8	MARK FOUND	1	1	1	559	null
new-coco-val2017-512	21,839	instagram	worst	overlay	0	reyes	{"filter":"reyes"}	11dc0f17ad20dd28bc13ac5433edef036a8e07dc1cbfb6e671258cc247f3e317	53bbcb7ba2d1dc908665f1a816aad54506bef6a32c7f8375712f2264c6a4c55d	NO MARK FOUND	0	0	0	651	null
new-coco-val2017-512	21,839	light_text	random	overlay	1	size59px-alpha157	{"text_px":59,"text_alpha":157,"text_at":[[176,105],[93,139],[53,164]],"text":["WATERMARK","sample text","hello world"]}	66931a3009c61802f3c0135c9f04a9bb13b59c55098166a2e38d18c6a86d53d3	98a6adbe4d1f5cecb9391618ee1ff972b9306da671235ef2b75565090b7f643d	MARK FOUND	1	1	1	506	null
new-coco-val2017-512	21,839	light_text	random	overlay	0	size59px-alpha157	{"text_px":59,"text_alpha":157,"text_at":[[176,105],[93,139],[53,164]],"text":["WATERMARK","sample text","hello world"]}	4539837b57f164867b7c5c3136b4f324c020e868304e4032257b9de8e065a5f7	86f478dfde5a6a693ff0a5280f5f637184f623474bf8210eeb6983f81a4acd17	NO MARK FOUND	0	0	0	600	null
new-coco-val2017-512	21,839	light_text	worst	overlay	1	size60px-alpha160	{"text_px":60,"text_alpha":160,"text_at":[[270,263],[67,273],[58,100]],"text":["WATERMARK","sample text","hello world"]}	d9f49778e6d6c106acf3946052d6f311a973bdce83adbd22c2c373a50a6bd834	6b7c6c3abc4458ddcb8f737e3e4d80476c5e0e7253f4b48ff5fe3a50cd11e8fa	MARK FOUND	1	1	1	489	null
new-coco-val2017-512	21,839	light_text	worst	overlay	0	size60px-alpha160	{"text_px":60,"text_alpha":160,"text_at":[[270,263],[67,273],[58,100]],"text":["WATERMARK","sample text","hello world"]}	93dfa732010502f27eb25efd9a75c223d2e24fc6b504151bfd7804f23bcfc2eb	0d4be8e0e58b2bd472332912ef44c8f22dd19074f3680fbe232c7a179eeaadc5	NO MARK FOUND	0	0	0	613	null
new-coco-val2017-512	21,839	file_format	random	quality	1	webp75	{"file_format":"webp75","webp_quality":75}	c8a091de6d7a425342d4d7a40d8fafc8e1c1f097852e56590b064b69762c8e90	b9d1d15cc547e1fd057d7459b325abd82031081132e6a14dabf969e1192a400b	MARK FOUND	1	1	1	485	null
new-coco-val2017-512	21,839	file_format	random	quality	0	webp75	{"file_format":"webp75","webp_quality":75}	d07802d2df831c5fb802d1edcf23a2d0e335f5568c82fd68be2ebd292b25806c	398903ac5553dcd9aa74dfa6747363316b619076de3d2f786fe20deab097d7e7	NO MARK FOUND	0	0	0	615	null
new-coco-val2017-512	21,839	file_format	worst	quality	1	webp50	{"file_format":"webp50","webp_quality":50}	cc50326690cf0f82eca44ccf8bf27e3cab61f164d051feb93e82aa0c9060b616	43586462c933ae2c8a79345275dab2c03730cf17023ee46d2271f0cda47d0edd	MARK FOUND	1	1	1	498	null
new-coco-val2017-512	21,839	file_format	worst	quality	0	webp50	{"file_format":"webp50","webp_quality":50}	a9a2666328599ccbcae7a4ef6e9b365156fdaa509b7ec3564a18c6711cec8766	246aea18fea7328334ba763574ce4d19391b7931b106601425c49b23a2da96e5	NO MARK FOUND	0	0	0	635	null
new-coco-val2017-512	21,839	gaussian_blur	random	quality	1	sigma0.926px	{"blur_sigma":0.9263355649310019}	cd773704125d257f688a56111ae8470b04df0da3d6ea0628fa1263f1d4aa5eb3	243c8d017a331dd09546dc2c92647608a98d6a8dde7a5814b9998f672a1771b4	MARK FOUND	1	1	1	535	null
new-coco-val2017-512	21,839	gaussian_blur	random	quality	0	sigma0.926px	{"blur_sigma":0.9263355649310019}	fc8431d16351d28f892a88bc0f0cc371e3eeb8d3cb7c247f7db3256dfdfd3250	9ba1c7f6b397553643a1386a641472944c44a8d9c8878001bf30aed13e2560bb	NO MARK FOUND	0	0	0	638	null
new-coco-val2017-512	21,839	gaussian_blur	worst	quality	1	sigma3px	{"blur_sigma":3.0}	e87a8f9a07fa732ff5196d373549d389ec1dcd06d56cf646bdd37d0d83b24b75	b9c19f7ce3e38108f46f24e404a533120859b0e7e4a94ffd3a483ff58377d706	MARK FOUND	1	1	1	514	null
new-coco-val2017-512	21,839	gaussian_blur	worst	quality	0	sigma3px	{"blur_sigma":3.0}	6f43bae527aaf7e322d0ac38bd86527647c38935e55a6e4b7a305eacbb16825e	d5ec54adbba5b5dc672d66140cad69c1309c670864dc3f1d65a8d225cf63d652	NO MARK FOUND	0	0	0	633	null
new-coco-val2017-512	21,839	denoise	random	quality	1	median7px	{"median_k":7}	c994ab47ab16216967bce2f676151d2478f58d118f8d45a32ccdc906217e1d0b	296b5e274ef82268006bd4fac6b0c10dcdd5f0fdea8d2388008c03eadf044709	MARK FOUND	1	1	1	471	null
new-coco-val2017-512	21,839	denoise	random	quality	0	median7px	{"median_k":7}	1bf8277b182676019acde68cccf34d633351a7f1d9ce749302328f41f59a4fe3	2aa1b43768d9dc13cb489cfd5057c6ea83d95fab727d9001225e15d49a4e5ec4	NO MARK FOUND	0	0	0	624	null
new-coco-val2017-512	21,839	denoise	worst	quality	1	median7px	{"median_k":7}	c994ab47ab16216967bce2f676151d2478f58d118f8d45a32ccdc906217e1d0b	296b5e274ef82268006bd4fac6b0c10dcdd5f0fdea8d2388008c03eadf044709	MARK FOUND	1	1	1	493	null
new-coco-val2017-512	21,839	denoise	worst	quality	0	median7px	{"median_k":7}	1bf8277b182676019acde68cccf34d633351a7f1d9ce749302328f41f59a4fe3	2aa1b43768d9dc13cb489cfd5057c6ea83d95fab727d9001225e15d49a4e5ec4	NO MARK FOUND	0	0	0	607	null
new-coco-val2017-512	21,839	crop_resize	random	spatial	1	area0.735	{"crop_area":0.735406068984136,"crop_box":[54,27,493,466]}	9805b505d3a7942d6b2797175b90735286425d6262f4aac03e3613879d2b8e43	e547b1bea98dfcbe59aa730ce0164f66d99e54dfd5d49dd8cf59e547a4f84355	MARK FOUND	1	1	1	440	null
new-coco-val2017-512	21,839	crop_resize	random	spatial	0	area0.735	{"crop_area":0.735406068984136,"crop_box":[54,27,493,466]}	54473825d11630315a58fda734249e92bc7eb91b0b052aeae5f0fba569fca7c5	0db413cdb492033abbdfd7d03345b3739ca0410e12ba5de9259db1e456154f51	NO MARK FOUND	0	0	0	615	null
new-coco-val2017-512	21,839	crop_resize	worst	spatial	1	area0.25	{"crop_area":0.25,"crop_box":[53,193,309,449]}	cd2fcf1aeb0b5f51a68651fbe1bd3e727d3dae1b07344f0863c0b47f3b69538b	9da0de9fce9c09ba449749bbdf9c11644c910953723359b715489a5b86c337a1	MARK FOUND	1	1	1	1,100	null
new-coco-val2017-512	21,839	crop_resize	worst	spatial	0	area0.25	{"crop_area":0.25,"crop_box":[53,193,309,449]}	b397b333fcd783d889720357d6a279fb56c508c4ae568184239c042b74bbb0ed	c72163e6de69104be47addbd3d96b06e2b85126ac004c22bdf7826dff9d04c34	NO MARK FOUND	0	0	0	619	null
new-coco-val2017-512	21,839	flip_lr	random	spatial	1	mirror	{"flip":"left-right"}	14c869785b77ed68b3bdb202adff2c350c082ae98b64d5f7abdcfd3dceb6c56c	4fba3ad4b0b80181823b2a326a6e08aadf600b85b56b58ca334de1a849c3fe42	MARK FOUND	1	1	1	506	null
new-coco-val2017-512	21,839	flip_lr	random	spatial	0	mirror	{"flip":"left-right"}	de23528018a330380e41009492fc5db6d0d3160fb0248a0ee7aa9ab4bd6de038	6d24751d9527a15b2630dfd0aafba21c932dcfd86f97d185c9b6260798d334c9	NO MARK FOUND	0	0	0	627	null
new-coco-val2017-512	21,839	flip_lr	worst	spatial	1	mirror	{"flip":"left-right"}	14c869785b77ed68b3bdb202adff2c350c082ae98b64d5f7abdcfd3dceb6c56c	4fba3ad4b0b80181823b2a326a6e08aadf600b85b56b58ca334de1a849c3fe42	MARK FOUND	1	1	1	489	null
new-coco-val2017-512	21,839	flip_lr	worst	spatial	0	mirror	{"flip":"left-right"}	de23528018a330380e41009492fc5db6d0d3160fb0248a0ee7aa9ab4bd6de038	6d24751d9527a15b2630dfd0aafba21c932dcfd86f97d185c9b6260798d334c9	NO MARK FOUND	0	0	0	622	null
new-coco-val2017-512	21,839	flip_ud	random	spatial	1	upside-down	{"flip":"top-bottom"}	2616f247dfceaa942fc1ea8b9e0a1a85f7dfcf5fda3e417c5ff9aadb756c2d07	3fb22d0719ffbc6a662e9aa5171c74b18feecceeff55f004bd761aad57f1ef8c	MARK FOUND	1	1	1	501	null
new-coco-val2017-512	21,839	flip_ud	random	spatial	0	upside-down	{"flip":"top-bottom"}	acc0e8ce01252a3d93d2e5b8e00217233bed7da01f09be5c696b27483bd5f9b6	0c904242df44b6de2091c7535378b45b7e3f11c673b73c0fdb9af1a29c0c7b86	NO MARK FOUND	0	0	0	638	null
new-coco-val2017-512	21,839	flip_ud	worst	spatial	1	upside-down	{"flip":"top-bottom"}	2616f247dfceaa942fc1ea8b9e0a1a85f7dfcf5fda3e417c5ff9aadb756c2d07	3fb22d0719ffbc6a662e9aa5171c74b18feecceeff55f004bd761aad57f1ef8c	MARK FOUND	1	1	1	525	null
new-coco-val2017-512	21,839	flip_ud	worst	spatial	0	upside-down	{"flip":"top-bottom"}	acc0e8ce01252a3d93d2e5b8e00217233bed7da01f09be5c696b27483bd5f9b6	0c904242df44b6de2091c7535378b45b7e3f11c673b73c0fdb9af1a29c0c7b86	NO MARK FOUND	0	0	0	649	null
new-coco-val2017-512	21,839	rotation	random	spatial	1	-15.06deg	{"rotate_deg":-15.055106928101122}	d808374cc99a9104a4d91754ceb220e6dc1b42ca83ae34bd78ad77d4ded316d5	f0ac3cda0e1a3527629d688510662782827a9eb05e46b1f2af33512929cfda6b	MARK FOUND	1	1	1	630	null
new-coco-val2017-512	21,839	rotation	random	spatial	0	-15.06deg	{"rotate_deg":-15.055106928101122}	58ede068fef4812d3b66425d492516de1dac1a66a45860efd6283321ae8e73ef	552e8bddbce32b36232834e9a8cd9b9681c91285d76938913eaa5648774cbd0b	NO MARK FOUND	0	0	0	673	null
new-coco-val2017-512	21,839	rotation	worst	spatial	1	45deg	{"rotate_deg":45}	d5b07cb8fb99cc67cff780cf01929ddf80fd9e832e74dd747da4813b6b9ba8a5	1fea24aa5dd7955afeecc339036f416cc1ed8ad342ee36bb1c432a71863bd20b	MARK FOUND	1	1	1	554	null
new-coco-val2017-512	21,839	rotation	worst	spatial	0	45deg	{"rotate_deg":45}	a3d14dd98292ce365175e77ddc86814948abc89c8a6c0296c4f34bd0f14e0c48	5cf31226c28ce88ee721cb4b91d8c05ebd07d8f0997a2f2c6e74ee673100cda7	NO MARK FOUND	0	0	0	601	null

End of preview. Expand in Data Studio

Vereye on the SynthID-Image robustness protocol: 72,000 blind owner-identification verdicts

On 300 COCO photos at 512 px never used in development, the Vereye invisible image watermark named the exact registered owner of 99.70 % of attacked copies at random strength and 95.80 % at the most severe strength, under the 29 transformations of Google DeepMind's SynthID-Image protocol, reading each copy blind, with its stronger mark for small images (PSNR 35.58 dB). An earlier reader, at standard strength and under a rule written before the run, identified 99.03 % / 86.38 % on 100 held-out photos; on 50 of those photos and the same attacks, Meta's Watermark Anything recovered its exact message on 80.41 % / 68.14 % and Adobe's TrustMark on 75.59 % / 62.97 %. No owner was named on any of the 36,000 identically attacked unmarked copies. Google reports 99.98 % / 99.72 % detection and 99.96 % / 98.84 % message recovery for SynthID-O on its own images, ahead of Vereye at both strengths.

Vereye is an early-stage startup founded by Adam Weisman. This benchmark was run by Vereye, so it is not independent, and it is not affiliated with or endorsed by Google DeepMind, Meta or Adobe. Every verdict is here, so anyone can re-score it.

Owner identified means the exact 64-bit key registered for the photo was read from the attacked copy alone, passed its 24-bit check, and was found in the registry. No score threshold is involved. The method and its limits are in the paper Identification, Not Detection: vereye.ai/paper.

Results

set	photos, side	mark strength (mean PSNR)	owner identified, random	owner identified, most severe	unmarked twins: owner named / any key read
unseen COCO val2017, current system (not pre-registered)	300, 512 px	35.58 dB (stronger mark for small images)	99.70 % (8,674 / 8,700)	95.80 % (8,335 / 8,700)	0 / 0 of 18,000
held-out COCO val2017, earlier reader, pre-registered	100, 512 px	41.6 dB (standard)	99.03 % (2,872 / 2,900)	86.38 % (2,505 / 2,900)	0 / 0 of 6,000
held-out DIV2K train, pre-registered	100, 1024 px	41.47 dB (standard)	99.90 % (2,897 / 2,900)	98.93 % (2,869 / 2,900)	0 / 0 of 6,000
development DIV2K valid, not evidence	100, 1024 px	41.41 dB (standard)	99.93 % (2,898 / 2,900)	99.48 % (2,885 / 2,900)	0 / 0 of 6,000

Conservative 95 % bounds for every rate, and a per-transformation table, are in package/results/summary.txt.

Rivals on the same 50 held-out COCO photos and the same seeded attacks (exact message decoded, each with its official code):

method	random	most severe
Vereye (exact owner)	99.17 %	86.21 %
Meta, Watermark Anything	80.41 %	68.14 %
Adobe, TrustMark	75.59 %	62.97 %

Unmarked copies: across all four sets, no owner was named, and no key passing the check was read, from any of the 36,000 attacked unmarked twins (46,467 across all of Vereye's benchmarks).

Read this before comparing

SynthID itself is not public, so it was not run. Google's figures are on its own 10,000 ImageNet images, at strengths it does not publish.
Two mark strengths. The pre-registered 512 px set used Vereye's standard strength (41.6 dB). The 300-photo unseen set was sealed with Vereye's design for small images, a stronger mark (35.58 dB): a 512 × 512 image holds a quarter of the mark of a 1024 × 1024 one.
Where the losses are: the 86.38 % most-severe average is pulled down mostly by one group of 4 of the 29 attacks, which combine crop, rotation, darkening and heavy JPEG in a single copy: 26.75 % (107 / 400) on the pre-registered set and 71.92 % (863 / 1,200) on the unseen set. The per-transformation table is in package/results/summary.txt.
The development set is not evidence: the writer and the reader were tuned on those photos. The held-out and unseen sets carry the claim.
1024 px is not the protocol's size. Only the 512 px sets compare with SynthID.
The strengths, the Instagram-style filters and the photo choice are ours. The SynthID-Image paper publishes none of them; ours are in package/scripts/transforms.py. Both rivals score above the values that paper reports for them, so our attacks are probably milder than Google's. That flatters every rate here, Vereye's included.
What cannot be checked yet: the sealed photos are withheld (below), so anyone can re-score these verdicts, but no one outside Vereye can yet re-read the attacked copies.
The record: package/HISTORY.md says what was tuned on what and which runs were stopped. The pre-registrations (package/PREREGISTRATION*.md) were written and frozen before the held-out runs. package/CRITICS.md lists every objection raised in an adversarial review and how each was settled, including the ones still open. These are working logs, kept verbatim.

Verify in three commands

hf download Vereye/benchmark-synthid-image-protocol --repo-type dataset --local-dir vereye-bench
cd vereye-bench
python3 package/scripts/score.py | cmp - package/results/summary.txt && echo "every number reproduced"

Python 3 standard library only: every number above is recomputed from the verdict tables, and the printout is identical, byte for byte, to package/results/summary.txt. To check the files themselves: cd package && sha256sum -c PUBLIC-SHA256SUMS.

Columns

cells.csv, one row per attacked copy and per unmarked twin:

column	meaning
`transformation`, `mode`, `params`	the attack, `random` or `worst` (most severe) strength, and the parameters drawn
`group`	the protocol's six groups, plus `extra`: our additional JPEG row, included in the unmarked-twin counts but not in the headline rates (hence 2,900 copies per strength, not 3,000)
`marked`	1 = attacked copy of the sealed photo; 0 = its identically attacked unmarked twin
`identified`	1 = the photo's registered key was read from the copy
`verdict`	the reader's text output
`owners_named` / `other_owners`	registered keys read / registered keys of other photos read (0 on every copy)
`crc_keys`	keys passing the 24-bit check, registered or not
`ms`	wall time of the benchmark harness per copy on a shared machine, including file I/O; not the product's speed
`pixel_sha256`, `file_sha256`, `file`	the copy's hashes and its path in the complete package (the image itself is not published)

rivals.csv: ident = the exact message decoded (the rates quoted above); bitacc, rawacc, mask, present = each rival's own scores, calibrated at 0.1 % false positives in summary.txt; psnr = the rival's mark strength.

What is in this dataset

path	content
`package/sets/<set>/results/cells.csv`	every verdict (columns above)
`package/sets/heldout-coco-val2017-512/results/rivals.csv`	Meta's and Adobe's watermarks on the same photos and attacks
`package/sets/<set>/photos/<id>/info.json`	per photo: the source, its hashes, the credit, the result summary
`package/results/`	the summary of every set
`package/datasets/`	the SHA-256 of every source photo, so the originals can be fetched from COCO and DIV2K and checked
`package/scripts/`	the preparation, attack, rival and scoring scripts (MIT)
`package/PREREGISTRATION*.md`, `HISTORY.md`, `CRITICS.md`	the decision rules, the history, the adversarial review
`package/PUBLIC-SHA256SUMS`	the hashes of every file published here (`sha256sum -c` passes)
`package/FILES.sha256`, `package/SHA256SUMS`	the file list and hashes of the complete package, including the withheld files, so a `sha256sum -c` against them reports those files as missing

What is not in this dataset

No images. The sealed photos are not published: many marked / unmarked pairs would let anyone study the mark. The originals are fetched from their own servers.
No decoder, and no Vereye code.
The frozen-code manifests are summarised to a file count and the SHA-256 of the full list.
Scripts that need the images (check.py, verify_on_demo.py and the others the package documents name) are in the complete package only.
The complete package is available on request, for research and evaluation. It adds the sealed photos, the full manifests, the original README, and the end-to-end verifier, which re-reads the shipped copies, rebuilds the unmarked twins from the public datasets, and can test any copy against the live demo. The SHA-256 of its sorted file list is fd5015c1e9b2ed7a88aa37df13e94d2b9bbd32937d4e35daa7b5f3d36ef0e0b5.

Citation

@misc{weisman2026vereyesynthid,
  author       = {Weisman, Adam},
  title        = {Vereye on the SynthID-Image Robustness Protocol: 72,000 Blind Owner-Identification Verdicts},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/Vereye/benchmark-synthid-image-protocol}}
}
@techreport{weisman2026identification,
  author      = {Weisman, Adam},
  title       = {Identification, Not Detection: Blind Owner Identification and Tamper Localization for Images with an Analytic False-Owner Bound},
  institution = {Vereye},
  type        = {Technical report, version 2},
  year        = {2026},
  url         = {https://vereye.ai/paper}
}

Credits and licences

Photos.

COCO val2017: Lin et al., Microsoft COCO: Common Objects in Context, ECCV 2014.
DIV2K: Agustsson and Timofte, NTIRE 2017 Challenge on Single Image Super-Resolution: Dataset and Study, CVPR Workshops 2017.
No photo is redistributed here, only hashes; copyright stays with the original owners.

Protocol. S. Gowal et al., SynthID-Image: Image watermarking at internet scale, arXiv:2510.09263, 2025.

Rivals. Watermark Anything (Meta) and TrustMark (Adobe), each run with its official code and published settings.

Licences. Verdict data and documents: CC BY 4.0. Scripts: MIT (package/LICENSE-scripts.txt).

Vereye, an early-stage startup founded by Adam Weisman · patents pending · adam@vereye.ai · vereye.ai · paper · live demo · press kit

Downloads last month: -

Paper for Vereye/benchmark-synthid-image-protocol

SynthID-Image: Image watermarking at internet scale

Paper • 2510.09263 • Published Oct 10, 2025 • 5