""" Draw multiple tracking results and annotations on video at the same time

Params:

videos: list of path to video directory (each directory contains video frames)

annots: list of path to annotation files

trk_results: list of list of path to tracking results

isMotformat: True, False, or 'DET'(corresponding results are detection results)

Return:

None, video will be stored in output_video

"""

video_id = 0

for video, annot in zip(videos, annots):

video_name = basename(video)

if basename(annot).find(video_name) == -1:

raise Exception("No corresding video and annotation!")

if video_name != "person14_2":

video_id += 1

continue

print("Processing " + video_name)

FPS = 30

# remember to modify frame width and height before testing video

frame_width = 1280

frame_height = 720

video_writer = cv2.VideoWriter("output_video/"+video_name+'.avi', cv2.VideoWriter_fourcc('M','J','P','G'), FPS, (frame_width, frame_height))

image_paths = sorted(glob.glob(os.path.join(video, '*jpg')))

sort_nicely(image_paths)

labels = np.loadtxt(annot, delimiter=',')

preds_per_trk = []

for trk_r in trk_results:

preds = np.loadtxt(trk_r[video_id], delimiter=',')

preds_per_trk.append(preds)

nb_lost = 0.0

for fi, image_path in enumerate(tqdm(image_paths)):

image = cv2.imread(image_path)

for trk_id, preds in enumerate(preds_per_trk):

# Iterate over tracker results

if isMotformat[trk_id] is True:

# Multi-object tracking results

index_list = np.argwhere(preds[:, 0] == (fi+1))

if index_list.shape[0] != 0:

max_iou = 0.0

target_index = -1

for index in index_list[:, 0]:

bbox = preds[index, 2:6]

iou = bbox_iou(bbox, labels[fi])

if iou > max_iou:

max_iou = iou

target_index = index

if max_iou > 0.2 and target_index != -1:

target_bbox = preds[target_index, 2:6]

target_bbox = xywh_to_xyxy(target_bbox)

cv2.rectangle(image,

(int(target_bbox[0]),int(target_bbox[1])),

(int(target_bbox[2]),int(target_bbox[3])),

colours[trk_id], 2)

cv2.rectangle(image,

(int(target_bbox[0])-2,int(target_bbox[1]-20)),

(int(target_bbox[0])+20+int(preds[target_index, 1])//10 ,int(target_bbox[1])+1),

colours[trk_id], -1)

cv2.putText(image,

str(int(preds[target_index, 1])),

(int(target_bbox[0]+2), int(target_bbox[1])-3),

cv2.FONT_HERSHEY_SIMPLEX,

0.6,

(0,0,0), 2)

else:

nb_lost += 1

elif isMotformat[trk_id] == 'DET':

# Detection results

index_list = np.argwhere(preds[:, 0] == (fi+1))

if index_list.shape[0] != 0:

max_iou = 0.0

target_index = -1

for index in index_list[:, 0]:

bbox = preds[index, 2:6]

iou = bbox_iou(bbox, labels[fi])

if iou > max_iou:

max_iou = iou

target_index = index

if max_iou > 0.5 and target_index != -1:

target_bbox = preds[target_index, 2:6]

target_bbox = xywh_to_xyxy(target_bbox)

cv2.rectangle(image,

(int(target_bbox[0]),int(target_bbox[1])),

(int(target_bbox[2]),int(target_bbox[3])),

colours[trk_id], 4)

else:

# Single object tracking results

rect = xywh_to_xyxy(preds[fi])

cv2.rectangle(image,

(int(rect[0]),int(rect[1])),

(int(rect[2]),int(rect[3])),

# (colours[trk_id]), 4)

(255,0,0), 4)

if isNAN(labels[fi]) is not True:

gt_rect = xywh_to_xyxy(labels[fi])

cv2.rectangle(image,

(int(gt_rect[0]),int(gt_rect[1])),

(int(gt_rect[2]),int(gt_rect[3])),

(0,0,255), 2)

video_writer.write(image)

# dir_path = 'output_video/person14_1/'

# frame_name = "{0:0=3d}".format(fi) + ".jpg"

# cv2.imwrite(dir_path + frame_name, image)

if fi > 180:

break

video_id += 1