Open sgbaird opened 2 years ago
@sgbaird Did you get anything? I am looking for regression for Video input (Batch_size, time_frames, Height, Width, Channels).
@ravikiranrao not yet, I think we're going to try a fully connected neural network to change the shape to 2D (and use PyTorch's built-in), but if we find something I'll try to remember to post back here.
Hi,
sorry for the late answer. M3D-Cam is designed for segmentation and classification, but it should also be adaptable to regression. Maybe it works out-of-the-box, but I have not tested that. However, you probably need to adapt some parts of the postprocessing in medcam_inject.py. Best would be for you to just try the default usage and work yourself up from that in case of errors. However, I won't be able to help you with that (if you try it) as I don't really have time currently.
# Import M3d-CAM
from medcam import medcam
# Init your model and dataloader
model = MyCNN()
data_loader = DataLoader(dataset, batch_size=1, shuffle=False)
# Inject model with M3d-CAM
model = medcam.inject(model, output_dir="attention_maps", save_maps=True)
# Continue to do what you're doing...
# In this case inference on some new data
model.eval()
for batch in data_loader:
# Every time forward is called, attention maps will be generated and saved in the directory "attention_maps"
output = model(batch)
# more of your code...
Best Karol
Any alternative codebases that might help out with regression? (instead of classification/segmentation).