search

ijkguo / mx-rcnn

Parallel Faster R-CNN implementation with MXNet.
Other
669 stars 292 forks source link

How to avoid using JPEG images in pascal format #103

Closed Ram-Godavarthi closed 5 years ago

Ram-Godavarthi commented 5 years ago

Hello Everyone,

I am currently using Pascal format dataset with jpeg images, annotations, imagesets.

I do not want to save the images as there will some loss of information when generating jpeg images in my application. Insead i wanna make use of Numpy array or other way to feed the input image data..

How can i do that? Where in the code is JPEG image is getting converted to numpy array..?

Can someone help me out in this. Where exactly i can feed the direct numpy array of input image instead of JPEG image..

ijkguo commented 5 years ago

JPEG image getting converted to numpy array symdata/image.py

Ram-Godavarthi commented 5 years ago

So in the first block, im_tensor is created for every image one by one ?? There i need to give the numpy array of each image?? How can i change the code to take the input as numpy array? I need something to feed all the data at once, Not by storing each numpy array and loading it everytime..

ijkguo commented 5 years ago

symdata/loader.py call symdata.image.get_image to read source image.