One Tap AI - BiRefNet Lite INT8 Tiled ONNX

This repository hosts the quantized, deform-tiled BiRefNet Lite INT8 ONNX model for high-resolution (1024x1024) on-device background removal in One Tap AI.

Model file: birefnet_lite_int8_tiled.onnx
Precision: INT8 (QDQ per-channel, Swin backbone kept fp32)
Size: ~167.4 MB (175,517,293 bytes)
Input: [1, 3, 1024, 1024]
Output: [1, 1, 1024, 1024]
License: MIT

Downloads last month: -; Downloads are not tracked for this model. How to track