New answers tagged

1 vote
Accepted

Is it bad to have similar images in training and validation set?

As in the comments: You're right, if the validation images are very similar to the training ones you may end up overfitting. But most of the times it is the best you can do so you do it. In a video, ...
user avatar
  • 226
0 votes

Intersection Over Union / Bounding Box for multiple bounding boxes in an image

Yes, normally IoU is calculated for all possible pairs within an image. Then the ones with the highest IoU score are selected. You can take a look at this code if you are not sure how to implement ...
user avatar
  • 1
1 vote
Accepted

How can I improve my current model to get a higher mAP value? (Stuck at 79~78)

There are several things to consider: Increase the number of convolution blocks. Use residual blocks, as explained here Use different activation functions, such as Leaky ReLU, Mish, Swish, etc. To ...
user avatar
1 vote
Accepted

Does the environment matter (area outside the box) in tensorflow's object detection algorithm?

The training is based on the boxes content only, but during the detection process, the algorithm has to scan all the image. Consequently, there is no learning of the environment outside the box. Such ...
user avatar
1 vote

How to retain the learnings from the original model when it is used as base model for training against new data?

When you use new data, the weights of the notwork change according to your new data to learn recognizing them , if you want to keep the weights related to old data as well, you should retrain the ...
user avatar
3 votes

What is difference and factors that help to make a decision between choosing training using pre-trained model vs from scratch?

Does pre-trained model in the above quote mean that it will re-use the training data from the pre-trained model (I can't see the original files used in the pre-trained model), plus my 300 images (100 ...
user avatar
  • 16k

Top 50 recent answers are included