def makeDir(imFolder, outFolder, labelFile, dirFile, classRef, res=(300, 300)):
coords, imNames, classes = get_labels(labelFile)
print('Got labels')
imFiles = listdir(imFolder)
with open(dirFile, 'w') as outFile:
for im in imFiles:
if im.endswith('.tif'):
arr = np.array(Image.open(path.join(imFolder, im)))
chips, boxes, boxClasses = chip_image(arr,
coords[imNames == im],
classes[imNames == im],
res)
for i in range(len(chips)):
chip = Image.fromarray(chips[i])
newName = path.join(outFolder,
"{}_{}.jpg".format(im[:-4], i))
chip.save(newName)
boxStr = []
emptyChip = False
for j in range(len(boxes[i])):
box = boxes[i][j]
boxClass = boxClasses[i][j]
if boxClass == 0:
emptyChip = True
classID = getRealID(int(boxClasses[i][j]), classRef)
boxStr.append("{},{},{},{},{}".format(
int(box[0]), int(box[1]), int(box[2]), int(box[3]),
classID))
if not emptyChip:
infoStr = '{} {}\n'.format(newName, " ".join(boxStr))
outFile.write(infoStr)
# print(newName)
print(im)
def chip_one_image(src, json_file):
img = Image.open(src)
arr = np.array(img)
coords, classes = get_labels_for_chip(src, json_file)
c_img, c_box, c_cls = wv.chip_image(img=arr, coords=coords, classes=classes, shape=(416, 416))
return c_img, c_box, c_cls
print(sd, flag1, flag2)
trn_path = Path(args.output) / 'chip_train/'
val_path = Path(args.output) / 'chip_valid/'
trn_path.mkdir(parents=True, exist_ok=True)
val_path.mkdir(parents=True, exist_ok=True)
resolution = [args.resolution] #416 # 672
## Train
out_path = trn_path
for idx in tqdm(idx_np[:trn_sz]):
fname = idx2fname[idx]
arr = wv.get_image(fname)
for res in resolution:
im, box, classes_final = wv.chip_image(
arr, coords[img_fn == fname.name],
classes[img_fn == fname.name], (res, res))
for idx, (x, b, c) in enumerate(
zip(im, box.values(), classes_final.values())):
if len(c) == 1 and c[0] == 0: continue ## no instance
else:
im_data = Image.fromarray(x)
tmp_fn = '_'.join(map(str, [fname.stem, idx, res]))
im_data.save(out_path / (tmp_fn + '.jpg'))
with open(str(out_path / (tmp_fn + '.txt')),
'w') as anno_f:
yolob = convert_to_yolo_box(x.shape[:2], b)
for i in range(len(c)):
anno_f.write(
str(idremap[int(c[i])]) + ' ' +
' '.join(map(str, yolob[i])) + '\n')
with open('xview_class_labels.txt') as f:
for row in csv.reader(f):
labels[int(row[0].split(":")[0])] = row[0].split(":")[1]
pass
pass
i = 0
for chip_name in tqdm.tqdm(all_images[:300]):
chip_name = (chip_name.split("/")[-1])
img_name = chip_name.split(".")[0]
coords = coords1[chips1 == chip_name]
classes = classes1[chips1 == chip_name].astype(np.int64)
c_img, c_box, c_cls = wv.chip_image(img=arr,
coords=coords,
classes=classes,
shape=(600, 600))
# xmin,ymin,xmax,ymax = c_box
# 0 1 2 3
for c_idx in (range(c_img.shape[0])):
# Save the chipped image
c_name = "{:06}_{:02}".format(int(img_name), c_idx)
# Save the chipped label
widths = c_box[c_idx][:, 2] - c_box[c_idx][:, 0]
x = c_box[c_idx][:, 0] + (widths / 2)
heights = c_box[c_idx][:, 3] - c_box[c_idx][:, 1]
y = c_box[c_idx][:, 1] + (heights / 2)
szx = c_img[c_idx].shape[0]
szy = c_img[c_idx].shape[1]
print(thechips[0])
print(theclasses[0])
#yo = chip_image('2355.tif', thecoords[0], theclasses[0])
chip_name = '104.tif'
arr = wv.get_image(chip_name)
plt.figure(figsize=(10,10))
plt.axis('off')
plt.imshow(arr)
print('\nyo\n')
c_img, c_box, c_cls = wv.chip_image(img = arr, coords=thecoords, classes=theclasses, shape=(1000,1000))
print("Num Chips: %d" % c_img.shape[0])
print("\nnumber of chips : ", c_img.shape[0])
w = c_img.shape[0]
#We can plot some of the chips
fig,ax = plt.subplots(3)
fig.set_figheight(5)
fig.set_figwidth(5)
for k in range(9):
plt.subplot(3,3,k+1)
plt.axis('off')
plt.imshow(c_img[np.random.choice(range(c_img.shape[0]))])
all_coords, all_chips, all_classes = wv.get_labels(OLD_ROOT +
"/xView_train.geojson")
max_gt_boxes = 0
# Iterate over each tif, and chip
for e, tif in enumerate(tif_list):
print("{} / {}".format(e, len(tif_list)))
# Get the info relevant to this single full frame .tif
ff_coords = all_coords[all_chips == tif]
ff_classes = all_classes[all_chips == tif].astype(np.int64)
# Chip the image into smaller pieces
arr = wv.get_image(OLD_ROOT + "/train_images/" + tif)
c_img, c_box, c_cls = wv.chip_image(img=arr,
coords=ff_coords,
classes=ff_classes,
shape=chip_shape)
num_chips = len(c_img)
# Iterate over chips for this tif
for i in range(num_chips):
# If chip has no gt boxes, skip
if len(c_box[i]) == 1 and c_box[i].all() == 0:
continue
if len(c_box[i]) > max_gt_boxes:
max_gt_boxes = len(c_box[i])
max_gt_chip_num = i
max_gt_tif = tif
print("new best:", max_gt_boxes, max_gt_tif, max_gt_chip_num)
print("max_gt_boxes:", max_gt_boxes)
xyz += 1
print("Working on: [{} / {}] {} ".format(xyz, len(tif_names), unique_tif))
# Make sure the file exists
if not os.path.isfile(OLD_ROOT + "/train_images/" + unique_tif):
continue
# Get the info relevant to this single full frame .tif
ff_coords = all_coords[all_chips == unique_tif]
ff_classes = all_classes[all_chips == unique_tif].astype(np.int64)
print("\tTotal Num Targets: ", len(ff_classes))
# Chip the image into smaller pieces
arr = wv.get_image(OLD_ROOT + "/train_images/" + unique_tif)
c_img, c_box, c_cls = wv.chip_image(img=arr,
coords=ff_coords,
classes=ff_classes,
shape=CHIP_SHAPE)
num_chips = len(c_img)
print("\tNum Chips: ", num_chips)
# For each image chip (i in range(num_chips))
for i in range(num_chips):
print("\t\tChip #: ", i)
print("\t\t\tNum Targets in Chip: ", len(c_cls[i]))
#print c_cls[i]
#print c_box[i]
# Git rid of very small boxes that are artifacts of chipping
final_boxes = []
final_classes = []
# Make sure the file exists
if not os.path.isfile(OLD_ROOT + "/train_images/" + unique_tif):
continue
# Get the info relevant to this single full frame .tif
ff_coords = all_coords[all_chips == unique_tif]
ff_classes = all_classes[all_chips == unique_tif].astype(np.int64)
#print("\tTotal Num Targets: ",len(ff_classes))
# Add multiscale loop
for scale in CHIP_SHAPE:
# Chip the image into smaller pieces
arr = wv.get_image(OLD_ROOT + "/train_images/" + unique_tif)
c_img, c_box, c_cls = wv.chip_image(img=arr,
coords=ff_coords,
classes=ff_classes,
shape=(scale, scale))
num_chips = len(c_img)
#print("\tNum Chips: ",num_chips)
# For each image chip (i in range(num_chips))
for i in range(num_chips):
#print("\t\tChip #: ",i)
# Calculate the center of the chip
center = (int(c_img[i].shape[0] / 2), int(c_img[i].shape[1] / 2))
# For each of the desired rotation degrees
for val in range(0, 360, 10):
