search

sejong-rcv / INSANet

INSANet: INtra-INter Spectral Attention Network for Effective Feature Fusion of Multispectral Pedestrian Detection
Creative Commons Attribution 4.0 International
17 stars 1 forks source link

Inquiry about Hyperparameters for Reproducing Experiments #4

Closed HuCutie closed 2 months ago

HuCutie commented 2 months ago

Hi, I am currently working on reproducing the experiments from your paper and your work. And I would like to ensure that my implementation closely matches the results reported. Could you please provide details on the hyperparameters used during the training process? I am using the recommended software environment and default parameters provided in the code. I have trained and tested the model on both single and dual A100 GPUs, but unfortunately, I am unable to achieve the results reported in your paper.

Specifically, I am interested in information regarding:

Learning rate Batch size Number of epochs Optimizer type and any associated parameters Any other relevant hyperparameters or settings Your assistance would be greatly appreciated, and it will be very helpful for the accuracy of my replication efforts.

Thank you very much for your time and support.

childult-programmer commented 2 months ago

Thanks for your interest and sorry for late reply.

Please see config.py for all hyperparameters.

Also, performance reported in paper is not the last (40) epoch, so if all settings are the same, we recommend evaluating each epoch to see the results.