ekami / carvana-challenge Goto Github PK

I tried using your code for random_shift_scale_rotate(..) but I can't see any rotation in the result.
I understand the code by I can't find why the problem is there.
Thanks.

Can you show us the steps to do inference using this model?

hello,
i have download carvana data set from kaggle i want to upload that manully please tell me how could i do this.
i am facing lots of errors during this.
i have read images from folder to code but i dont know what other changes i have to do so please tell me that

reading of images code is

directory_input_images = "./train_hq/"
directory_input_lable = "./train_masks/"

im_array = []
no_of_im = 0
files = glob.glob (directory_input_images+"*.jpg")
for myFile in files:
	img = cv2.imread(myFile, 0)
	img = cv2.resize(img, (572, 572))
	im_array.append (img)
	#im_array = np.reshape(im_array, (-1,200,200,1)) 


	no_of_im = no_of_im + 1
	#return no_of_im

error i am facing is

raise TypeError("filename must be a str or bytes object, or a file")
TypeError: filename must be a str or bytes object, or a file

Hi,

Could you please add a license file to your repo so others may benefit from your code?

Thanks

First of all, thanks for sharing this well written solution!

For an image of size (N, N). The following line of code seems to return an array of size (N,), but the desired effect is to classify every pixel in that image, that is, an array of size(N, N):

So maybe change it to pred_arr = pred.data.cpu().numpy() would make more sense.

(I use pytorch 0.2.0.post2)