I've created a model with Auto ML from Google Cloud Platform. This model has 5 labels. I uploaded more than 1000 images for each label. 400 picture for the label None_of_the_above. I trained the machine with all uploaded pictures. Record created for multi-label scenario.
When I predict a picture with multiple objects after training, objects are not recognized correctly. However, when I predict images with a single image, mostly correct values are returned.
Can someone help me to understand what I am doing wrong here?