-
Notifications
You must be signed in to change notification settings - Fork 31
/
get_pitch_frames.py
196 lines (158 loc) · 7.35 KB
/
get_pitch_frames.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
import colorsys
import copy
import random
import time
import cv2
import numpy as np
import tensorflow as tf
from PIL import Image
from scipy.ndimage import shift
from src.FrameInfo import FrameInfo
from src.utils import distance, fill_lost_tracking
from src.generate_overlay import generate_overlay, draw_ball_curve
from src.SORT_tracker.sort import Sort
from src.SORT_tracker.tracker import Tracker
# Get the pitching section in the whole video
def get_pitch_frames(video_path, infer, input_size, iou, score_threshold):
print("Video from: ", video_path)
vid = cv2.VideoCapture(video_path)
width = int(vid.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(vid.get(cv2.CAP_PROP_FRAME_HEIGHT))
fps = int(vid.get(cv2.CAP_PROP_FPS))
track_colors = [(161, 235, 52), (83, 254, 92), (255, 112, 52), (161, 235, 52), (255, 235, 52), (255, 38, 38), (255, 235, 52), (210, 235, 52), (52, 235, 131), (52, 64, 235), (0, 0, 255), (0, 255, 255),
(255, 0, 127), (127, 0, 127), (255, 127, 255), (127, 0, 255), (255, 255, 0), (255, 0, 0), (0, 0, 255), (0, 255, 0), (0, 255, 255), (255, 0, 255), (50, 100, 150), (10, 50, 150), (120, 20, 220)]
# Store the pitching section in pitch_frames
pitch_frames = []
detected_balls = []
tracked_balls = []
frames = []
tracker_min_hits = 3
frame_id = 0
# Create Object Tracker
tracker = Sort(max_age=8, min_hits=tracker_min_hits, iou_threshold=0.3)
while True:
return_value, frame = vid.read()
if return_value:
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
frames.append(FrameInfo(frame, False))
else:
if frame_id == vid.get(cv2.CAP_PROP_FRAME_COUNT):
print("Processing complete")
break
raise ValueError("Something went wrong! Only MP4 format is accepted.")
# Detect the baseball in the frame
detections = detect(infer, frame, input_size, iou, score_threshold, detected_balls)
# Feed in detections to obtain SORT tracking
if(len(detections) > 0):
trackings = tracker.update(np.array(detections))
else:
trackings = tracker.update()
# Add the valid trackings to balls_list
for t in trackings:
t = [int(i) for i in t]
start = (t[0], t[1])
end = (t[2], t[3])
# cv2.rectangle(frame, start, end, (255, 0, 0), 5)
# cv2.putText(frame, str(t[4]), start, cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 255, 255), 2, cv2.LINE_AA)
color = track_colors[t[4] % 12]
centerX = int((t[0] + t[2]) / 2)
centerY = int((t[1] + t[3]) / 2)
tracked_balls.append([centerX, centerY, color])
# Store the frames with ball tracked
if(len(trackings) > 0):
# Only run at the first track from SORT
if(len(pitch_frames) == 0):
last_tracked_frame = frame_id
add_balls_before_SORT(frames, detected_balls, tracked_balls, tracker_min_hits)
# Add prior 20 frames before the first balsadl
pitch_frames.extend(frames[-20:])
# Add lost frames if any
add_lost_frames(frame_id, last_tracked_frame, frames, pitch_frames)
# Append the frame with detected ball location
last_ball = tuple(tracked_balls[-1][:-1])
pitch_frames.append(FrameInfo(frame, True, last_ball, color))
last_tracked_frame = frame_id
result = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR)
detection = cv2.resize((result), (0, 0), fx=0.5, fy=0.5)
cv2.imshow("result", detection)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
frame_id += 1
# Use Polyfit to approximate the untracked balls
fill_lost_tracking(pitch_frames)
# Add five more frames after the last tracked frame
pitch_frames.extend(frames[last_tracked_frame: last_tracked_frame+5])
return pitch_frames, width, height, fps
# Tensorflow Object Detection API Sample
def detect(infer, frame, input_size, iou, score_threshold, detected_balls):
image_data = cv2.resize(frame, (input_size, input_size))
image_data = image_data / 255.
image_data = image_data[np.newaxis, ...].astype(np.float32)
batch_data = tf.constant(image_data)
pred_bbox = infer(batch_data)
for key, value in pred_bbox.items():
boxes = value[:, :, 0:4]
pred_conf = value[:, :, 4:]
boxes, scores, classes, valid_detections = tf.image.combined_non_max_suppression(
boxes=tf.reshape(boxes, (tf.shape(boxes)[0], -1, 1, 4)),
scores=tf.reshape(
pred_conf, (tf.shape(pred_conf)[0], -1, tf.shape(pred_conf)[-1])),
max_output_size_per_class=50,
max_total_size=50,
iou_threshold=iou,
score_threshold=score_threshold
)
boxes = boxes.numpy()
scores = scores.numpy()
classes = classes.numpy()
valid_detections = valid_detections.numpy()
offset = 30
accuracyThreshold = 0.95
frame_h, frame_w, _ = frame.shape
detections = []
for i in range(valid_detections[0]):
score = scores[0][i]
if(score > accuracyThreshold):
coor = boxes[0][i]
coor[0] = (coor[0] * frame_h)
coor[2] = (coor[2] * frame_h)
coor[1] = (coor[1] * frame_w)
coor[3] = (coor[3] * frame_w)
centerX = int((coor[1] + coor[3]) / 2)
centerY = int((coor[0] + coor[2]) / 2)
print(f'Baseball Detected ({centerX}, {centerY}), Confidence: {str(round(score, 2))}')
# cv2.circle(frame, (centerX, centerY), 15, (255, 0, 0), -1)
detected_balls.append([centerX, centerY])
detections.append(np.array([coor[1]-offset, coor[0]-offset, coor[3]+offset, coor[2]+offset, score]))
return detections
def add_balls_before_SORT(frames, detected, tracked, tracker_min_hits):
distance_threshold = 100
first_ball = tracked[0]
color = first_ball[2]
balls_to_add = []
# Get the untracked balls that's close enough to the first tracked ball
for untracked in detected[-(tracker_min_hits+1):]:
if(distance(untracked, first_ball) < distance_threshold):
untracked.append(color)
balls_to_add.append(untracked)
# Add the untracked balls to frame
modify_frames = frames[-(tracker_min_hits+1):]
balls_to_add_temp = copy.deepcopy(balls_to_add)
for point in balls_to_add_temp:
del point[2]
balls_to_add_temp = np.array(balls_to_add_temp, dtype='int32')
for idx, frame in enumerate(modify_frames):
# cv2.polylines(frame.frame, [balls_to_add_temp[:idx+1]], False, color, 22, lineType=cv2.LINE_AA)
frames[-((tracker_min_hits+1)-idx)] = FrameInfo(frame.frame, True, tuple(balls_to_add_temp[idx]), color)
def add_lost_frames(frame_id, last_tracked_frame, frames, pitch_frames):
if(frame_id - last_tracked_frame > 1):
print('Lost frames:', frame_id - last_tracked_frame)
frames_to_add = frames[last_tracked_frame: frame_id]
# Mark the detection in lost in this frame
for ball_frame in frames_to_add:
ball_frame.ball_lost_tracking = True
pitch_frames.extend(frames_to_add)
# def get_bright_color():
# h, s, l = random.random(), 0.5 + random.random()/2.0, 0.4 + random.random()/5.0
# r, g, b = [int(256*i) for i in colorsys.hls_to_rgb(h, l, s)]
# return [r, g, b]