Buckets:
| { | |
| "stats": { | |
| "mix_path": "/workspace/input/LLaVA-Tuning/llava_v1_5_mix665k.json", | |
| "image_root": "/workspace/input/LLaVA-Tuning", | |
| "seed": 42, | |
| "n_samples": 30, | |
| "entries_total": 665297, | |
| "entries_with_image": 624609, | |
| "entries_excluded_as_evaluation_images": 10712, | |
| "eligible_candidates": 613897, | |
| "missing_image_files_skipped": 0, | |
| "evaluation_image_keys_indexed": 13278 | |
| }, | |
| "samples": [ | |
| { | |
| "sample_id": "000000335492", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000335492.jpg", | |
| "relative_image": "coco/train2017/000000335492.jpg", | |
| "prompt": "What is the main object in the image?" | |
| }, | |
| { | |
| "sample_id": "000000346082", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000346082.jpg", | |
| "relative_image": "coco/train2017/000000346082.jpg", | |
| "prompt": "What is unique about the woman's appearance in the image?" | |
| }, | |
| { | |
| "sample_id": "000000253789", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000253789.jpg", | |
| "relative_image": "coco/train2017/000000253789.jpg", | |
| "prompt": "What is the color of the cat in the image?" | |
| }, | |
| { | |
| "sample_id": "000000389041", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000389041.jpg", | |
| "relative_image": "coco/train2017/000000389041.jpg", | |
| "prompt": "How many adult elephants are visible in the image?" | |
| }, | |
| { | |
| "sample_id": "000000516198", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000516198.jpg", | |
| "relative_image": "coco/train2017/000000516198.jpg", | |
| "prompt": "How many people are in the image?" | |
| }, | |
| { | |
| "sample_id": "000000440811", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000440811.jpg", | |
| "relative_image": "coco/train2017/000000440811.jpg", | |
| "prompt": "Can you elaborate on the elements of the picture provided?" | |
| }, | |
| { | |
| "sample_id": "000000472922", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000472922.jpg", | |
| "relative_image": "coco/train2017/000000472922.jpg", | |
| "prompt": "Explain the visual content of the image in great detail." | |
| }, | |
| { | |
| "sample_id": "000000505111", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000505111.jpg", | |
| "relative_image": "coco/train2017/000000505111.jpg", | |
| "prompt": "What factors should one consider while choosing the perfect meal from this table?" | |
| }, | |
| { | |
| "sample_id": "000000086032", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000086032.jpg", | |
| "relative_image": "coco/train2017/000000086032.jpg", | |
| "prompt": "What safety concerns could arise from the man's actions?" | |
| }, | |
| { | |
| "sample_id": "000000061439", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000061439.jpg", | |
| "relative_image": "coco/train2017/000000061439.jpg", | |
| "prompt": "What potential challenges or risks do these skateboarders face while practicing their tricks in this particular setting?" | |
| }, | |
| { | |
| "sample_id": "000000530551", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000530551.jpg", | |
| "relative_image": "coco/train2017/000000530551.jpg", | |
| "prompt": "What potential safety concerns should drivers be aware of in this setting?" | |
| }, | |
| { | |
| "sample_id": "000000265074", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000265074.jpg", | |
| "relative_image": "coco/train2017/000000265074.jpg", | |
| "prompt": "How many different vegetables are there? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "000000530834", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000530834.jpg", | |
| "relative_image": "coco/train2017/000000530834.jpg", | |
| "prompt": "Is this a new bus? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "000000224047", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000224047.jpg", | |
| "relative_image": "coco/train2017/000000224047.jpg", | |
| "prompt": "Is the man wearing wristband? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "000000073584", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000073584.jpg", | |
| "relative_image": "coco/train2017/000000073584.jpg", | |
| "prompt": "Is this a forest? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "000000346862", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000346862.jpg", | |
| "relative_image": "coco/train2017/000000346862.jpg", | |
| "prompt": "What is the child wearing? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "000000519685", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000519685.jpg", | |
| "relative_image": "coco/train2017/000000519685.jpg", | |
| "prompt": "What are they standing on? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "000000302383", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000302383.jpg", | |
| "relative_image": "coco/train2017/000000302383.jpg", | |
| "prompt": "What does the information on the blue legged placard describe? A. bus B. motorcycle C. menu D. protest Answer with the option's letter from the given choices directly." | |
| }, | |
| { | |
| "sample_id": "000000008458", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000008458.jpg", | |
| "relative_image": "coco/train2017/000000008458.jpg", | |
| "prompt": "Please provide the bounding box coordinate of the region this sentence describes: the woman in the back wearing a blue bikini top." | |
| }, | |
| { | |
| "sample_id": "000000157341", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000157341.jpg", | |
| "relative_image": "coco/train2017/000000157341.jpg", | |
| "prompt": "Please provide the bounding box coordinate of the region this sentence describes: right thing." | |
| }, | |
| { | |
| "sample_id": "000000340317", | |
| "image_path": "/workspace/input/LLaVA-Tuning/coco/train2017/000000340317.jpg", | |
| "relative_image": "coco/train2017/000000340317.jpg", | |
| "prompt": "Please provide the bounding box coordinate of the region this sentence describes: woman in white skirt." | |
| }, | |
| { | |
| "sample_id": "VG_100K/1159841", | |
| "image_path": "/workspace/input/LLaVA-Tuning/vg/VG_100K/1159841.jpg", | |
| "relative_image": "vg/VG_100K/1159841.jpg", | |
| "prompt": "Please provide the bounding box coordinate of the region this sentence describes: Red curb along sidewalk." | |
| }, | |
| { | |
| "sample_id": "VG_100K_2/2404861", | |
| "image_path": "/workspace/input/LLaVA-Tuning/vg/VG_100K_2/2404861.jpg", | |
| "relative_image": "vg/VG_100K_2/2404861.jpg", | |
| "prompt": "Please provide a short description for this region: [0.69, 0.39, 0.98, 0.87]." | |
| }, | |
| { | |
| "sample_id": "VG_100K/2323588", | |
| "image_path": "/workspace/input/LLaVA-Tuning/vg/VG_100K/2323588.jpg", | |
| "relative_image": "vg/VG_100K/2323588.jpg", | |
| "prompt": "Please provide the bounding box coordinate of the region this sentence describes: A shed is behind the train." | |
| }, | |
| { | |
| "sample_id": "503", | |
| "image_path": "/workspace/input/LLaVA-Tuning/gqa/images/503.jpg", | |
| "relative_image": "gqa/images/503.jpg", | |
| "prompt": "Is there any cup to the right of the napkin that is on top of the tray? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "1572841435", | |
| "image_path": "/workspace/input/LLaVA-Tuning/ocr_vqa/images/1572841435.jpg", | |
| "relative_image": "ocr_vqa/images/1572841435.jpg", | |
| "prompt": "Who wrote this book? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "743202953", | |
| "image_path": "/workspace/input/LLaVA-Tuning/ocr_vqa/images/743202953.jpg", | |
| "relative_image": "ocr_vqa/images/743202953.jpg", | |
| "prompt": "Who wrote this book? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "61452971", | |
| "image_path": "/workspace/input/LLaVA-Tuning/ocr_vqa/images/61452971.jpg", | |
| "relative_image": "ocr_vqa/images/61452971.jpg", | |
| "prompt": "Who wrote this book? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "1419701886", | |
| "image_path": "/workspace/input/LLaVA-Tuning/ocr_vqa/images/1419701886.jpg", | |
| "relative_image": "ocr_vqa/images/1419701886.jpg", | |
| "prompt": "Who wrote this book? Answer the question using a single word or phrase." | |
| }, | |
| { | |
| "sample_id": "0102022aa14f4c76", | |
| "image_path": "/workspace/input/LLaVA-Tuning/textvqa/train_images/0102022aa14f4c76.jpg", | |
| "relative_image": "textvqa/train_images/0102022aa14f4c76.jpg", | |
| "prompt": "Provide a one-sentence caption for the provided image. Reference OCR token: FUJFILM, 0, X-T1, LocK.pr" | |
| } | |
| ] | |
| } |
Xet Storage Details
- Size:
- 9.1 kB
- Xet hash:
- 470cfb79fc39136a8c48514bfa43a2c7a6a4da86e6c84f4c7231d6e5e3c5643e
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.