search

alexgkendall / caffe-segnet

Implementation of SegNet: A Deep Convolutional Encoder-Decoder Architecture for Semantic Pixel-Wise Labelling
http://mi.eng.cam.ac.uk/projects/segnet/
Other
1.08k stars 452 forks source link

Using Custom Data #45

Open eweill opened 8 years ago

eweill commented 8 years ago

I have trained the SegNet network and have a resulting Caffe model. I would like to now test it on some of my own data.

Is there a simple way to test the model using a custom data set (I am unable to locate in the code where the images are being taken from for the testing and display in test_segmentation_camvid.py)

Thanks.

Timo-hab commented 8 years ago

Siehe #18

eweill commented 8 years ago

Thanks, I think I have it working now.

Timo-hab commented 8 years ago

@eweill How good are your results with your own images? I rescaled my images to 460x360, but the results are not so well as in CamVid-Dataset. In one case, an RGB image was better in the other case, the grayscale image.

eweill commented 8 years ago

@Timo-hab My results are definitely not at the same level as those in the CamVid-Dataset. I am getting alot of areas within the images where it is not detecting objects at all or there are just many different segmentation patterns because it is unable to decipher what an actual object is. I am currently doing more testing to see if it is my images or how I trained the network. I am also looking into using the webdemo model to see if that helps at all.

I have not tested on grayscale image either, but I will definitely test and let you know how well it goes.

Nirmitee commented 7 years ago

@eweill nad @Timo-hab I am also using the segnet_basic model on my own dataset, but not able to get good resullts. The training accuracy is good but the predicted images are vague. Also I am not achieving good global accuracy too. Can you suggest me something, so that the images and global accuracy improve?