fcjian / TOOD

TOOD: Task-aligned One-stage Object Detection, ICCV2021 Oral

fcjian/TOOD Issues

TAL样本分配策略与损失函数计算是否会造成网络输出的类别置信度偏低
Updated 3 months ago
T-Head
Updated 10 months ago
Welcome update to OpenMMLab 2.0
Updated a year ago
Tood's onnx file request
Updated a year ago3
Plot result
Updated a year ago10
N个连续的卷积层
Updated 2 years ago8
TOOD
Updated 2 years ago
是否可以提供没有使用tap的tal训练config？
Updated 2 years ago
Layer Attention instead of Channel Attention?
Closed 2 years ago5
Benchmark / FPS
Updated 2 years ago
How to apply TOOD in YOLOX
Updated 2 years ago1
the image size
Updated 2 years ago1
Error during training (Assertion input_val >= zero && input_val <= one failed.)
Updated 2 years ago6
RuntimeError
Updated 2 years ago
How about the ATSS assigner as initial static assignment method?
Closed 2 years ago2
i changed the number of ratios, then model can not train ,where should i have to modify futher?
Updated 2 years ago
Support for Swin backbone
Updated 2 years ago
If I want to load the resnet50-19c8e357.pth from my local path instead of downloading from the terminal,how can I change the code?
Updated 3 years ago2
Can I use it in two-stage method?
Updated 3 years ago
RuntimeError
Updated 3 years ago3
a little puzzled about the T-Head module
Updated 3 years ago1
Can I ask some questions about TOOD/mmdet/core/bbox/assigners/task_aligned_assigner.py ?
Updated 3 years ago4
请问TAL中，t的计算公式是否为：t = (s**alpha) * (u**beta)，t = t/t.max * u.max？
Closed 3 years ago2
how to stead of deform_sampling?
Updated 3 years ago1
Coupled Head
Closed 3 years ago2
model weight
Updated 3 years ago4