Abe

what if images are different size each iteration?

3

Thank you for the great work. I'm currently working on ViT and had a question. It seems that the code is based on fixed input image size. How would the...

The code is written w.r.t batch size of 128. If i'm using say batch of 32, do i need to change lr accordingly? ex. 0.1 (batch=128) --> 0.025 (batch=32) Thank...

Zoom in and label

4

@AlexeyAB Is there a way to zoom in and label the object? I'm trying to label small objects as well as big objects and was wondering if there's a way...

network can't train when incorporate this

5

i have added perceiver to my current network and it seems like network can't be trained. AP is zero all the way and doesn't train at all. Does the code...

I saw that you used optical flow to determine the scene is static or not. What was the threshold value you used? Thank you!

Loss graph during training

41

Is there a way to show loss graph during training like tensorflow?

Question about equation

1

In equation 4,  what is the meaning of P(w,h)-1 part? why are you subtracting 1? Thank you in advance

It seems like this module is not trained with mixed precision. Do you plan on adding this feature? Thank you,

Quick question regarding inverted residual block. if expand_ratio == 1: self.conv = nn.Sequential( # dw nn.Conv2d(hidden_dim, hidden_dim, 3, stride, 1, groups=hidden_dim, bias=False), nn.BatchNorm2d(hidden_dim), nn.ReLU6(inplace=True), # pw-linear nn.Conv2d(hidden_dim, oup, 1, 1,...

When set num_heads = 1, doesnt train

12

First of all, thank you for the wonderful work. I have a quick question though. I see you have set num_head = 6, meaning you iteratively update bounding box results...