search

JiaxiongQ / DeepLiDAR

Deep Surface Normal Guided Depth Prediction for Outdoor Scene from Sparse LiDAR Data and Single Color Image (CVPR 2019)
MIT License
248 stars 49 forks source link

The application of ‘get_transform’ functions #20

Open gujiaqivadin opened 4 years ago

gujiaqivadin commented 4 years ago

processed = preprocess.get_transform(augment=False)

Hello, JiaxiongQ! Thanks for your excellent code. When I read you code in trainloader.py, I found you apply processed = preprocess.get_transform(augment=False) before each input(sparse, image, mask, and so). And I found that this function is used for making the input from numpy array format into Tensor format and normalize the whole input into 0~1 by dividing it by 256. But I am confused whether it is a must, will it influence the final output result?

JiaxiongQ commented 4 years ago

Sorry, we didn't do this experiment. But we don't normalize the sparse depth and mask, the depth was divided by 256 to get the true depth from the depth image, and the gt depth did the same operation.