I’m handling a detection problem with image coordinate (x,y with range -1 to 1) as a label.
The model is something like this :
class BB_model(nn.Module): def __init__(self): super(BB_model, self).__init__() resnet = models.resnet34(pretrained=False) layers = list(resnet.children())[:8] self.features1 = nn.Sequential(*layers[:6]) self.features2 = nn.Sequential(*layers[6:]) self.bb = nn.Sequential(nn.BatchNorm1d(512), nn.Linear(512, 2)) def forward(self, x): x = self.features1(x) x = self.features2(x) x = F.relu(x) x = nn.AdaptiveAvgPool2d((1,1))(x) x = x.view(x.shape, -1) x = self.bb(x) return x
L1loss as the loss function and
MAE as the metrics. Somehow the model is not learning anything…
Do I have to use tanh / sigmoid too?