quocthangnguyen / deep-metric-learning-tsinghua-dogs Goto Github PK

@QuocThangNguyen Hello. I am having a bit problem in running prepare_TsinghuaDogs.py script for dataset preparation and getting error IndexError: list index out of range. Just wanted to check with you if I am doing things correctly.

So in data folder I opened TsinghuaDogs where I opened the following folders
High-Annotations
high-resolution
TrainAndValList
train
val

Is this a correct data directory structure before running a prepare_TsinghuaDogs.py ?

@QuocThangNguyen Hello. I tried training and inference, they work fine (just training takes too long). Then I wanted to evaluate a trained model to see confusion matrix, T-SNE and similarity matrix, but I receive errors

FileNotFoundError: [WinError 3] The system cannot find the path specified: 'data/TshinghuaDogs/train'
FileNotFoundError: [WinError 3] The system cannot find the path specified: 'data/TshinghuaDogs/val'

I followed the advice how to structure dataset for train and val folders, so I don't know why it can't find directories.
Can you please help to figure out what is the issue and let me know if you need additional information from my side!

Hi,

I'm interested in using your dataset. However, when prepare_TsinghuaDogs.py is unzipping the downloaded files, an exception occurs:

Archive: data/TsinghuaDogs.zip
End-of-central-directory signature not found. Either this file is not
a zipfile, or it constitutes one disk of a multi-part archive. In the
latter case the central directory and zipfile comment will be found on
the last disk(s) of this archive.
unzip: cannot find zipfile directory in one of data/TsinghuaDogs.zip or
data/TsinghuaDogs.zip.zip, and cannot find data/TsinghuaDogs.zip.ZIP, period.
22-Jun-08 17:47:02 root INFO: Downloading https://cg.cs.tsinghua.edu.cn/ThuDogs/high-annotations.zip to data/TsinghuaDogs/high-annotations.zip

Could you please fix it or tell me the proper way to use it?

@QuocThangNguyen First of all, thanks for the great work on fine-grained classification. I would like to test the repository train and inference, but unfortunatly I could not find proxynca-resnet50.pth pre-trained model in checkpoints folder. Can you please let me know where I can find and dowload it. Look forward to hearing frolm you!

Dear QuocThangNguyen,

Thank you for your great work about Tsinghua Dogs dataset, I really like it.

In the main pape, you provide pretrained model, would you mind provide the usuage codes for the pretrained model? Did you do any preprocessing before fitting the image to the model?

You mention about other benchmarks in project main page, but there is not pretrained model, so I do not know the pretrained model in github page is using which method.

Thank you for your help.

Best Wishes,

Alex

@QuocThangNguyen Hello!

I would like to convert a pytorch (pth) trained model to TensorFlow Lite (tflite) format. Do you have some advice how to do it properly for your Fine-Grained classification models? I saw that in general the converting process has several steps, convert Pytorch to ONNX, ONNX to TensorFlow and then TensorFlow to TensorFlow Lite. But if you know a quick way to do it or maybe there is a special script that can do all these steps without errors, that would be great.

Look forward to hearing from you!

Dear QuocThangNguyen,

Thank you for your great work about Tsinghua Dogs dataset, I really like it.

In the main pape, you provide pretrained model, this model only provide the image embedding for retrieval rather than output the dog category. Is there a way to get the dog category given an image?

Thank you for your help.

Best Wishes,

Alex