pbar = progressbar.ProgressBar(maxval=conf["num_distraction_images"],
widgets=widgets).start()
print("[INFO] describing distraction ROIs...")
# loop over the desired number of distraction images
for i in np.arange(0, conf["num_distraction_images"]):
# randomly select a distraction images, load it, convert it to grayscale, and
# then extract random pathces from the image
print(dstPaths)
image = cv2.imread(random.choice(dstPaths))
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
patches = extract_patches_2d(
image,
tuple(conf["window_dim"]),
max_patches=conf["num_distractions_per_image"])
# loop over the patches
for patch in patches:
# extract features from the patch, then update teh data and label list
features = hog.describe(patch)
data.append(features)
labels.append(-1)
# update the progress bar
pbar.update(i)
# dump the dataset to file
pbar.finish()
print("[INFO] dumping features and labels to file...")
dataset.dump_dataset(data, labels, conf["features_path"], "features")
(boxes, probs) = od.detect(gray,
conf["window_dim"],
winStep=conf["hn_window_step"],
pyramidScale=conf["hn_pyramid_scale"],
minProb=conf["hn_min_probability"])
# loop over the bounding boxes
for (prob, (startX, startY, endX, endY)) in zip(probs, boxes):
# extract the ROI from the image, resize it to a known, canonical size, extract
# HOG features from teh ROI, and finally update the data
roi = cv2.resize(gray[startY:endY, startX:endX],
tuple(conf["window_dim"]),
interpolation=cv2.INTER_AREA)
features = hog.describe(roi)
data.append(np.hstack([[prob], features]))
# update the progress bar
pbar.update(i)
# sort the data points by confidence
pbar.finish()
print("[INFO] sorting by probability...")
data = np.array(data)
data = data[data[:, 0].argsort()[::-1]]
# dump the dataset to file
print("[INFO] dumping hard negatives to file...")
dataset.dump_dataset(data[:, 1:], [-1] * len(data),
conf["features_path"],
"hard_negatives",
writeMethod="a")