Pixel2Mesh
Pixel2Mesh copied to clipboard
Thanks for sharing the code! why only use 1 batch during training?
Thanks for sharing this great work! I am new to this field and have one question: why only use 1 batch during training?
Hello, as mentioned in walsvid/Pixel2MeshPlusPlus#2 the use of batch_size=1 is because we continue to use Pixel2Mesh's data processing method. The number of ground truth points obtained by sampling is inconsistent. For simplicity, we use batch_size=1. We have found a new ground truth data processing method, which will be updated later. If we can guarantee the same length of ground truth data, we can increase the batch.
@walsvid Did you update your ground truth data processing method?
And I have found you use queue with 64 maximum-size for data fetch. Why did you choose this when your batch size is only 1. https://github.com/nywang16/Pixel2Mesh/blob/003bb83a4211954125bcda7d5fe54cea6b6aec8d/p2m/fetcher.py#L28
The current release code is still the old processing method, because the length of gt is different, so it is still batch_size = 1. The new processing method will be updated with the new version code.