mihaidusmanu
commented
4 years ago Hello. As mentioned in the README, this repository is a reimplementation in PyTorch of the initial TensorFlow code used to train the d2_tf*.pth models. However, throughout my previous experiments (on PyTorch 1.0 mostly), the results obtained when retraining with the code contained in this repository were quite close on HPatches Sequences to the ones of the released model. Thus, I suspect there might have been some recent changes that affected convergence. Could you provide me with the set of weights obtain after retraining? Also, could you share the pdf with the HPatches Sequences results after retraining?
Extra question just to make sure: are you using the latest version of the code (with the updated dataset undistortion & preprocessing)?
liwenwei123
Dear Mihai Dusmanu! Thank you for your answer last time! I have run the code you provided , after training the default 10 epoch, i got the model saved in 'checkpoints/d2.best.pth', and replace the model downloaded in 'models/d2_tf.pth' in extract_feature.py. However,when i want to test the training result in HPatches-Sequences-Matching-Benchmark.ipynb,the feature and mathes i compute by the model in d2.best.pth is less than the ones in d2-net-trained precomputed errors.When i run Qualitative-Matches.ipynb to see the result,i found the plot result is also not as good as the old model:d2_tf.pth. i am wondering if there is any step i missed? How can i train the same result as model in d2_tf.pth? Thanks Sincerely!