JingruiYu
JingruiYu
I encountered the same problem, thank you for sharing!
但我在其他图片上跑出的结果也有较大差异。。。所以我想会不会是c++和python之间导致的精度下降？或者是什么其他原因？
  我是使用的这两张图片，使用python-v0.1.3的code和当前repo的master的code。谢谢！
更新一下，我发现原因应该是python code已经是1.0的版本了，c++这边对应的应该是v0.1.3？然后那边放出来的模型不同的名称也会导致结果不同。如果python用的也是v0.1.3，然后也是相同的onnx模型，那差异是很小的。不过从release的模型名字来看似乎只有resolution的不同吧，为什么会导致输出结果也不同呢？
谢谢还有个细节麻烦您也留意一下， 在[main.cpp line 48](https://github.com/OroChippw/LightGlue-OnnxRunner/blob/9fdb260525cd29e52460e6ef0494fede3ddf42a9/src/main.cpp#L48C44-L48C57)处有将BGR转RGB的，在[src/transform.cpp line16](https://github.com/OroChippw/LightGlue-OnnxRunner/blob/9fdb260525cd29e52460e6ef0494fede3ddf42a9/src/transform.cpp#L16C47-L16C47)还有一次BGR2RGB，且随后的[src/transform.cpp line95](https://github.com/OroChippw/LightGlue-OnnxRunner/blob/9fdb260525cd29e52460e6ef0494fede3ddf42a9/src/transform.cpp#L95)是BGR2GRAY。我想这预处理中应该有重复和错位的。比如只需要一次BGR2GRAY?或者最后的这里应该是RGB2GRAY？
Hello, have you successfully fine-tuned DPVO? I have tried fine-tuning it, but the [results](https://github.com/princeton-vl/DPVO/issues/91) are quite strange. Would you be willing to give me some guidance?
> I trained it on my own data, nan appears when I didn't normalize my gt data.maybe you can try normalizing everything to tartanair's distribution? > […](#) Wow, that's sound...
> tartan's range is about -10 to 10. I think the absolute value of gt should be less than 20 ...? > […](#) Thanks for your kindly repley. I checked...
Hello, have you successfully fine-tuned DPVO? I have tried fine-tuning it, but the [results](https://github.com/princeton-vl/DPVO/issues/91) are quite strange. Would you be willing to give me some guidance?
Hello, have you successfully fine-tuned DPVO? I have tried fine-tuning it, but the [results](https://github.com/princeton-vl/DPVO/issues/91) are quite strange. Would you be willing to give me some guidance?