Search results

不同backbone该如何设置self.lateral_channels_in_collection参数

Status: Open.

#283 In ZhengPeng7/BiRefNet;

· pedroHuang123 opened on Jan 31, 2026

多次调用backbone提取多尺度特征

Status: Open.

#282 In ZhengPeng7/BiRefNet;

· pedroHuang123 opened on Jan 27, 2026

deformable卷积在高通芯片不支持npu推理

Status: Open.

#281 In ZhengPeng7/BiRefNet;

· pedroHuang123 opened on Jan 23, 2026

Does the network support multi-class classification?

Status: Open.

#280 In ZhengPeng7/BiRefNet;

· hcscysh opened on Jan 19, 2026

TR-Human 数据集

Status: Open.

#279 In ZhengPeng7/BiRefNet;

· jxncyym opened on Jan 15, 2026

train

Status: Open.

#278 In ZhengPeng7/BiRefNet;

· zaoruis opened on Jan 14, 2026

Question: Performance drop when using DINOv3 as backbone on COD task

Status: Open.

#274 In ZhengPeng7/BiRefNet;

· YogurtFy opened on Dec 21, 2025

关于dinov3的模型训练

Status: Open.

#273 In ZhengPeng7/BiRefNet;

· chenCHEN-12 opened on Nov 28, 2025

train_question

Status: Open.

#272 In ZhengPeng7/BiRefNet;

· hcscysh opened on Nov 25, 2025

How can I convert this model to the ncnn format? Directly converting the compiled ONNX model using pnnx throws an exception.

Status: Open.

#269 In ZhengPeng7/BiRefNet;

· pangxiaobin opened on Oct 31, 2025

Make it easier to know which datasets were used

documentationImprovements or additions to documentationImprovements or additions to documentation

TODO

Status: Open.

#268 In ZhengPeng7/BiRefNet;

· set-soft opened on Oct 24, 2025

关于视频抠图的人像描边问题

Status: Open.

#264 In ZhengPeng7/BiRefNet;

· YanCeyMichael opened on Oct 13, 2025