mlboydaisuke/DIS-ISNet-ExecuTorch

dis_isnet — ExecuTorch

• Source: xuebinqin/DIS + NimaBoscarino/IS-Net_DIS-general-use weights
• License: Apache-2.0
• Input: [[1, 3, 1024, 1024]] — RGB, scaled to [-1,1] (x/255 then (x-0.5)/0.5), 1024x1024
• Output: alpha mask [1,1,1024,1024] 0-1 (sigmoid)

Variants

All variants take and return fp32 tensors — swap the .pte file, keep your app code.

build	file	size (MB)	parity vs fp32 eager (worst corr)	Mac median (ms)*
fp32	dis_isnet_xnnpack_fp32.pte	176.1	1.000000	123.4
Core ML (fp16, iOS)	dis_isnet_coreml_all.pte	89.0	0.999984	29.2

The Core ML build is the same graph lowered to Apple's Neural Engine instead of XNNPACK, which is CPU-only. Measured on an iPhone 17 Pro across seven models, it runs 3.5x to 13.9x faster (median 12x) at roughly half the file size — for example Depth-Anything-V2-Small at 500.8 ms against 42.7 ms, and MODNet at 81.7 ms against 5.9 ms. It computes in fp16 and is iOS-only; the XNNPACK files stay the portable option and are what runs on Android.

*Mac arm64, single process, median of 10 — a reference point for relative cost only, not a device number (torch eager fp32 on the same machine: 364.6 ms).

Builds that did not earn a slot

• fp16 is not shipped: worst-output corr 0.986 against fp32 eager, below the 0.995 bar for this precision. The file converts and runs; the numbers do not hold up, so it is left out rather than shipped with a warning.
• int8 is not shipped: measured in the units that matter for this model — mask IoU at 0.5: median 0.9106 over 10 real images, worst 0.4647.

Verification (executorch 1.4.0, torch 2.13.0)

Parity is measured against the fp32 eager model on real image input; corr is the correlation over all elements of each output tensor.

output	shape	max_abs_diff	corr
0	[1, 1, 1024, 1024]	3.606e-06	1.000000

XNNPACK delegate coverage (fp32): 100.0% (467/467 ops)

Conversion

torch.export -> to_edge_transform_and_lower(partitioner) -> .pte (conversion scripts: executorch-models)