-
### 📚 The doc issue
1. 我的标注数据只有4个点，使用rtmpose-m_8xb256-420e_coco-256x192.py的配置文件。我的标注图像是1920*1200的图像。训练时报错，我不清楚这是哪里出现了错误。
```
09/21 10:19:33 - mmengine - INFO -
-------------------------------…
-
看了下中文caption的评测程序，发现产生的句子首先经过jieba中文分词了，然后在计算四个指标前又经过 PTBTokenizer 分词了，请问这样做有什么原因吗？为什么不直接只做一步分词？
-
Hi MMPose Team,
I would like to ask a question about training a 2D pose model with combined datasets COCO, AI, OccHuman, etc.
There would be a much higher amount of training data by combining thes…
-
The current aic result is about 30 mAP both top-down and bottom-up methods. I think the evaluation may exist some bugs.
-
ZeroDivisionError: division by zero
Hi Dr. Luo. I try to evaluate model, but I meet an error.
error info:
```shell
python eval.py --model data/pretrain/fc/model-best.pth --infos_path data/pretr…
-
The cf_SPnet_aichallenger_224_128_0.54G_1.2 pre-quantized model available in quantized/Edge directory does not compile.
Model: https://www.xilinx.com/bin/public/openDownload?filename=cf_SPnet_aicha…
-
-
-
-
你好,
训练代码目前没有处理未标记关节点(x=0, y=0, v=3), 请问这个对模型结果会有影响么? 为什么不尝试推断关节的可见状态, 比如将v的状态embedd一起训练?