Regarding hyperparameters for training on UCF-QNRF dataset

cvlab-stonybrook / DM-Count

Code for NeurIPS 2020 paper: Distribution Matching for Crowd Counting.

MIT License

218 stars 52 forks source link

Regarding hyperparameters for training on UCF-QNRF dataset #20

Closed svshivapuja closed 3 years ago

svshivapuja commented 3 years ago

Hi, Firstly I would like to say that your paper and code are very impressive. Thank you for releasing the code.

The default hyperparameters worked well for NWPU and ShanghaiTech datasets but are giving a bad results (MAE -- 600) for UCF-QNRF. Could you please clarify about this ?

Thanks in advance :)

Boyu-Wang commented 3 years ago

Default hyperparameters should work for QNRF dataset as well. Since you are able to make it work for NWPU and ShanghaiTech dataset, you may need to check your data in QNRF. Make sure you preprocess the data. MAE of 600 is too large. You may also need to check the network output. Make sure it does not always output 0.

svshivapuja commented 3 years ago

Okay ! thanks :)