R, G, B = ([], [], [])
## loop over datasets
for dtype, dinfo in datasets.items():
paths, labels, hdfpath = dinfo
print("[INFO] building dataset = ", dtype, ", labels distribution =",
Counter(labels))
# build a hdf writer
writer = HDF5DatasetWriter(hdfpath, (len(paths), 256, 256, 3))
# loop over the image paths & preprocess images
for i in tqdm(range(len(paths))):
path, label = paths[i], labels[i]
image = cv2.imread(path)
image = aap.preprocess(image)
# save mean values of RGB channels for train set
if dtype == "train":
b, g, r = cv2.mean(image)[:3]
R.append(r)
G.append(g)
B.append(b)
# feed image & label to writer buffer
writer.add([image], [label])
pass
# close writer
writer.close()
pass
predictions = []
submission = pd.read_csv("./sample_submission.csv") # columns = [id,label]
# preprocess batch images & do prediction
if useTTA == "True":
print("[INFO] applying TTA..")
else:
print("[INFO] NOT applying TTA..")
# loop over batches
for i in tqdm(range(0, N, B)):
batchPaths = imagePaths[i:i + B]
batchImages = []
for path in batchPaths:
image = cv2.imread(path)
image = aap.preprocess(image) # maintain AR and resize to 256 x 256
image = iap.preprocess(image)
# Special for ImageNet dataset => substracting mean RGB pixel intensity
image = imagenet_utils.preprocess_input(image) # (256, 256, 3)
batchImages.append(image)
pass
if useTTA == "True":
for image in batchImages:
crops = cp1.preprocess(image)
# predict over 10 crops
crops_probs = model.predict(crops)
# predict probs of dogs
pred = crops_probs.mean(axis=0)[1]