search

bmoore20 / habs

Detect Harmful Algal Blooms (HABs) in images of the Finger Lakes.
0 stars 0 forks source link

Implement classifier functionality #8

Open bmoore20 opened 3 years ago

bmoore20 commented 3 years ago

Execute the functionality of HABs_Classify.py (classifying/predicting images with trained model, sorting classified images into correct folders, and generating a report) from original, but refactor it so that it fits into new organization of the repo.

This step also might involve reaching out to Professor Halfman and Professor Dumitriu from HWS to ask for feedback on their user experience with the original program, and to see if they have any recommendations for changes or improvements to the functionality.

bmoore20 commented 3 years ago

Talked to Professor Dumitriu over the weekend (3.21.21) and she said that her nor Professor Halfman have used my original program. Therefore, there is no available feedback about how to improve the user experience.

bmoore20 commented 3 years ago

Thought: Adjust HABsdataset so that it can take in video as well as images? This would be beneficial for when we want to classify the HABs with our trained model and don't need labels for the images (mode=classify). We can just feed in the video and have it tell us what times or frames the HABs are. Would save us the step of having to use VLC Media Player to extract images from the videos since the current images we are getting are from video.

bmoore20 commented 3 years ago

Figure out how to get the predictions array from PyTorch!

bmoore20 commented 3 years ago

https://stackoverflow.com/questions/51041128/pytorch-predict-single-example https://discuss.pytorch.org/t/making-a-prediction-with-a-trained-model/2193 https://discuss.pytorch.org/t/efficient-method-to-gather-all-predictions/8008 https://pytorch.org/docs/stable/generated/torch.max.html