Gabriel

Try installing the requirements.txt (although the versions are just going to be latest). I've run the same code and it has no problems: ```py import torch mask_encoder = torch.hub.load('milesial/Pytorch-UNet', 'unet_carvana',...

Not sure if that's what you want, but you can probably do the following ```py class MyEfficientNet(nn.Module): def __init__(self): super().__init__() # EfficientNet self.network = EfficientNet.from_pretrained("efficientnet-b0") # Replace last layer self.network._fc...

> > Not sure if that's what you want, but you can probably do the following > > > > ```python > > class MyEfficientNet(nn.Module): > > > > def...

As you first pre-train the model with an autoencoder, you have pretty good representations in the latent space. When you start training the Encoder only part, you could find two...

I don't understand the question properly, but in my experience, this error occurs when opencv expects a numpy array, and most of the time its receiving None instead of a...

The VGG part is encoding the features only, so the difference between the two here (by your knowledge of classifiers I believe) is that resnets have skip connections all the...

I would also remind you that ResNets use skip connections via addition (and not concatenation like the unet), so the idea of using the skip connections are purely to preserve...

I have this same issue, when training say Yolox_s, it uses 10.3gb vram no matter what, batch size doesnt matter, image size doesnt matter, it takes 10.3 and goes. Did...

nvm just found out that removing `-o` from training call shows whats being used, and that `-o` is to occupy the gpu mem ```py # yolox.utils.metric.py def occupy_mem(cuda_device, mem_ratio=0.9): """...

I believe the idea here is to detect things out of the supervised environment. In that case, object detection (as a supervised method) doesnt fit in such a task. Just...