Update app.py
Browse files
app.py
CHANGED
@@ -10,13 +10,13 @@ import os
|
|
10 |
# Load the trained YOLOv8n model
|
11 |
model = YOLO("best.pt")
|
12 |
|
13 |
-
# Constants
|
14 |
-
STUMPS_WIDTH = 0.2286 # meters
|
15 |
-
BALL_DIAMETER = 0.073 # meters
|
16 |
FRAME_RATE = 30 # Input video frame rate
|
17 |
-
SLOW_MOTION_FACTOR = 6
|
18 |
-
CONF_THRESHOLD = 0.3
|
19 |
-
|
20 |
|
21 |
def process_video(video_path):
|
22 |
if not os.path.exists(video_path):
|
@@ -25,30 +25,21 @@ def process_video(video_path):
|
|
25 |
frames = []
|
26 |
ball_positions = []
|
27 |
debug_log = []
|
28 |
-
frame_count = 0
|
29 |
-
max_frames = FRAME_RATE * 3 # Limit to 3 seconds of frames
|
30 |
|
31 |
-
|
|
|
32 |
ret, frame = cap.read()
|
33 |
if not ret:
|
34 |
break
|
35 |
frame_count += 1
|
36 |
-
|
37 |
-
|
38 |
-
frames.append(frame.copy()) # Store original frame
|
39 |
-
# Detect ball
|
40 |
-
results = model.predict(frame_resized, conf=CONF_THRESHOLD, imgsz=RESIZE_DIM)
|
41 |
detections = 0
|
42 |
-
scale_x, scale_y = frame.shape[1] / RESIZE_DIM, frame.shape[0] / RESIZE_DIM
|
43 |
for detection in results[0].boxes:
|
44 |
-
if detection.cls == 0: #
|
45 |
detections += 1
|
46 |
x1, y1, x2, y2 = detection.xyxy[0].cpu().numpy()
|
47 |
-
|
48 |
-
x1, x2 = x1 * scale_x, x2 * scale_x
|
49 |
-
y1, y2 = y1 * scale_y, y2 * scale_y
|
50 |
-
ball_center = [(x1 + x2) / 2, (y1 + y2) / 2]
|
51 |
-
ball_positions.append(ball_center)
|
52 |
cv2.rectangle(frame, (int(x1), int(y1)), (int(x2), int(y2)), (0, 255, 0), 2)
|
53 |
frames[-1] = frame
|
54 |
debug_log.append(f"Frame {frame_count}: {detections} ball detections")
|
@@ -63,105 +54,66 @@ def process_video(video_path):
|
|
63 |
|
64 |
def estimate_trajectory(ball_positions, frames):
|
65 |
if len(ball_positions) < 2:
|
66 |
-
return
|
|
|
|
|
|
|
67 |
x_coords = [pos[0] for pos in ball_positions]
|
68 |
y_coords = [pos[1] for pos in ball_positions]
|
69 |
times = np.arange(len(ball_positions)) / FRAME_RATE
|
70 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
71 |
try:
|
72 |
fx = interp1d(times, x_coords, kind='linear', fill_value="extrapolate")
|
73 |
fy = interp1d(times, y_coords, kind='quadratic', fill_value="extrapolate")
|
74 |
except Exception as e:
|
75 |
-
return
|
76 |
-
|
77 |
-
# Interpolate for all frames and future projection
|
78 |
-
t_all = np.linspace(0, times[-1] + 0.5, len(frames) + 10)
|
79 |
-
x_all = fx(t_all)
|
80 |
-
y_all = fy(t_all)
|
81 |
-
trajectory = list(zip(x_all, y_all))
|
82 |
-
return trajectory, t_all, "Trajectory estimated successfully"
|
83 |
-
|
84 |
-
def detect_impact_point(ball_positions, frames):
|
85 |
-
if len(ball_positions) < 3:
|
86 |
-
return ball_positions[-1] if ball_positions else None, len(ball_positions) - 1
|
87 |
-
# Assume batsman is near stumps (bottom center of frame)
|
88 |
-
frame_height, frame_width = frames[0].shape[:2]
|
89 |
-
batsman_x = frame_width / 2
|
90 |
-
batsman_y = frame_height * 0.8 # Approximate batsman position
|
91 |
-
min_dist = float('inf')
|
92 |
-
impact_idx = len(ball_positions) - 1
|
93 |
-
impact_point = ball_positions[-1]
|
94 |
-
|
95 |
-
# Look for sudden change in trajectory or proximity to batsman
|
96 |
-
for i in range(1, len(ball_positions) - 1):
|
97 |
-
x, y = ball_positions[i]
|
98 |
-
prev_x, prev_y = ball_positions[i-1]
|
99 |
-
next_x, next_y = ball_positions[i+1]
|
100 |
-
# Check direction change (simplified)
|
101 |
-
dx1, dy1 = x - prev_x, y - prev_y
|
102 |
-
dx2, dy2 = next_x - x, next_y - y
|
103 |
-
angle_change = abs(np.arctan2(dy2, dx2) - np.arctan2(dy1, dx1))
|
104 |
-
dist_to_batsman = np.sqrt((x - batsman_x)**2 + (y - batsman_y)**2)
|
105 |
-
if angle_change > np.pi/4 or dist_to_batsman < frame_width * 0.1:
|
106 |
-
impact_idx = i
|
107 |
-
impact_point = ball_positions[i]
|
108 |
-
break
|
109 |
-
|
110 |
-
return impact_point, impact_idx
|
111 |
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
frame_height, frame_width = frames[0].shape[:2]
|
119 |
-
stumps_x = frame_width / 2
|
120 |
-
stumps_y = frame_height * 0.9
|
121 |
-
stumps_width_pixels = frame_width * (STUMPS_WIDTH / 3.0)
|
122 |
-
|
123 |
-
pitch_point = ball_positions[0]
|
124 |
-
impact_point, impact_idx = detect_impact_point(ball_positions, frames)
|
125 |
-
|
126 |
-
# Check pitching point
|
127 |
-
pitch_x, pitch_y = pitch_point
|
128 |
-
if pitch_x < stumps_x - stumps_width_pixels / 2 or pitch_x > stumps_x Moderation: x > stumps_x + stumps_width_pixels / 2:
|
129 |
-
return f"Not Out (Pitched outside line at x: {pitch_x:.1f}, y: {pitch_y:.1f})", trajectory, pitch_point, impact_point
|
130 |
-
|
131 |
-
# Check impact point
|
132 |
-
impact_x, impact_y = impact_point
|
133 |
-
if impact_x < stumps_x - stumps_width_pixels / 2 or impact_x > stumps_x + stumps_width_pixels / 2:
|
134 |
-
return f"Not Out (Impact outside line at x: {impact_x:.1f}, y: {impact_y:.1f})", trajectory, pitch_point, impact_point
|
135 |
-
|
136 |
-
# Check trajectory hitting stumps
|
137 |
-
for x, y in trajectory:
|
138 |
-
if abs(x - stumps_x) < stumps_width_pixels / 2 and abs(y - stumps_y) < frame_height * 0.1:
|
139 |
-
return f"Out (Ball hits stumps, Pitch at x: {pitch_x:.1f}, y: {pitch_y:.1f}, Impact at x: {impact_x:.1f}, y: {impact_y:.1f})", trajectory, pitch_point, impact_point
|
140 |
return f"Not Out (Missing stumps, Pitch at x: {pitch_x:.1f}, y: {pitch_y:.1f}, Impact at x: {impact_x:.1f}, y: {impact_y:.1f})", trajectory, pitch_point, impact_point
|
141 |
|
142 |
-
def generate_slow_motion(frames, trajectory, pitch_point, impact_point,
|
143 |
if not frames:
|
144 |
return None
|
145 |
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
|
146 |
out = cv2.VideoWriter(output_path, fourcc, FRAME_RATE / SLOW_MOTION_FACTOR, (frames[0].shape[1], frames[0].shape[0]))
|
147 |
|
148 |
-
for
|
149 |
-
# Draw trajectory
|
150 |
-
|
151 |
-
|
152 |
-
|
153 |
|
154 |
-
# Draw pitch point
|
155 |
-
if pitch_point
|
156 |
x, y = pitch_point
|
157 |
-
cv2.circle(frame, (int(x), int(y)), 8, (0, 0, 255), -1)
|
158 |
cv2.putText(frame, "Pitch Point", (int(x) + 10, int(y) - 10),
|
159 |
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 0, 255), 2)
|
160 |
|
161 |
-
# Draw impact point
|
162 |
-
if impact_point
|
163 |
x, y = impact_point
|
164 |
-
cv2.circle(frame, (int(x), int(y)), 8, (0, 255, 255), -1)
|
165 |
cv2.putText(frame, "Impact Point", (int(x) + 10, int(y) + 20),
|
166 |
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 255), 2)
|
167 |
|
@@ -174,12 +126,11 @@ def drs_review(video):
|
|
174 |
frames, ball_positions, debug_log = process_video(video)
|
175 |
if not frames:
|
176 |
return f"Error: Failed to process video\nDebug Log:\n{debug_log}", None
|
177 |
-
trajectory,
|
178 |
-
decision, trajectory, pitch_point, impact_point = lbw_decision(ball_positions, trajectory, frames)
|
179 |
-
_, impact_idx = detect_impact_point(ball_positions, frames)
|
180 |
|
181 |
output_path = f"output_{uuid.uuid4()}.mp4"
|
182 |
-
slow_motion_path = generate_slow_motion(frames, trajectory, pitch_point, impact_point,
|
183 |
|
184 |
debug_output = f"{debug_log}\n{trajectory_log}"
|
185 |
return f"DRS Decision: {decision}\nDebug Log:\n{debug_output}", slow_motion_path
|
@@ -193,6 +144,7 @@ iface = gr.Interface(
|
|
193 |
gr.Video(label="Very Slow-Motion Replay with Ball Detection (Green), Trajectory (Blue), Pitch Point (Red), Impact Point (Yellow)")
|
194 |
],
|
195 |
title="AI-Powered DRS for LBW in Local Cricket",
|
|
|
196 |
)
|
197 |
|
198 |
if __name__ == "__main__":
|
|
|
10 |
# Load the trained YOLOv8n model
|
11 |
model = YOLO("best.pt")
|
12 |
|
13 |
+
# Constants for LBW decision and video processing
|
14 |
+
STUMPS_WIDTH = 0.2286 # meters (width of stumps)
|
15 |
+
BALL_DIAMETER = 0.073 # meters (approx. cricket ball diameter)
|
16 |
FRAME_RATE = 30 # Input video frame rate
|
17 |
+
SLOW_MOTION_FACTOR = 6 # For very slow motion (6x slower)
|
18 |
+
CONF_THRESHOLD = 0.3 # Confidence threshold for detection
|
19 |
+
IMPACT_ZONE_Y = 0.85 # Fraction of frame height where impact is likely (near stumps)
|
20 |
|
21 |
def process_video(video_path):
|
22 |
if not os.path.exists(video_path):
|
|
|
25 |
frames = []
|
26 |
ball_positions = []
|
27 |
debug_log = []
|
|
|
|
|
28 |
|
29 |
+
frame_count = 0
|
30 |
+
while cap.isOpened():
|
31 |
ret, frame = cap.read()
|
32 |
if not ret:
|
33 |
break
|
34 |
frame_count += 1
|
35 |
+
frames.append(frame.copy())
|
36 |
+
results = model.predict(frame, conf=CONF_THRESHOLD)
|
|
|
|
|
|
|
37 |
detections = 0
|
|
|
38 |
for detection in results[0].boxes:
|
39 |
+
if detection.cls == 0: # Assuming class 0 is the ball
|
40 |
detections += 1
|
41 |
x1, y1, x2, y2 = detection.xyxy[0].cpu().numpy()
|
42 |
+
ball_positions.append([(x1 + x2) / 2, (y1 + y2) / 2])
|
|
|
|
|
|
|
|
|
43 |
cv2.rectangle(frame, (int(x1), int(y1)), (int(x2), int(y2)), (0, 255, 0), 2)
|
44 |
frames[-1] = frame
|
45 |
debug_log.append(f"Frame {frame_count}: {detections} ball detections")
|
|
|
54 |
|
55 |
def estimate_trajectory(ball_positions, frames):
|
56 |
if len(ball_positions) < 2:
|
57 |
+
return None, None, None, "Error: Fewer than 2 ball detections for trajectory"
|
58 |
+
frame_height = frames[0].shape[0]
|
59 |
+
|
60 |
+
# Extract x, y coordinates
|
61 |
x_coords = [pos[0] for pos in ball_positions]
|
62 |
y_coords = [pos[1] for pos in ball_positions]
|
63 |
times = np.arange(len(ball_positions)) / FRAME_RATE
|
64 |
|
65 |
+
# Find impact point (closest to batsman, near stumps)
|
66 |
+
impact_idx = None
|
67 |
+
for i, y in enumerate(y_coords):
|
68 |
+
if y > frame_height * IMPACT_ZONE_Y: # Ball is near stumps/batsman
|
69 |
+
impact_idx = i
|
70 |
+
break
|
71 |
+
if impact_idx is None:
|
72 |
+
impact_idx = len(ball_positions) - 1 # Fallback to last detection
|
73 |
+
|
74 |
+
impact_point = ball_positions[impact_idx]
|
75 |
+
|
76 |
+
# Use positions up to impact for interpolation
|
77 |
+
x_coords = x_coords[:impact_idx + 1]
|
78 |
+
y_coords = y_coords[:impact_idx + 1]
|
79 |
+
times = times[:impact_idx + 1]
|
80 |
+
|
81 |
try:
|
82 |
fx = interp1d(times, x_coords, kind='linear', fill_value="extrapolate")
|
83 |
fy = interp1d(times, y_coords, kind='quadratic', fill_value="extrapolate")
|
84 |
except Exception as e:
|
85 |
+
return None, None, None, f"Error in trajectory interpolation: {str(e)}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
|
87 |
+
# Project trajectory (detected + future)
|
88 |
+
t_full = np.linspace(times[0], times[-1] + 0.5, len(times) + 10)
|
89 |
+
x_full = fx(t_full)
|
90 |
+
y_full = fy(t_full)
|
91 |
+
trajectory = listagus = f"Out (Ball hits stumps, Pitch at x: {pitch_x:.1f}, y: {pitch_y:.1f}, Impact at x: {impact_x:.1f}, y: {impact_y:.1f})", trajectory, pitch_point, impact_point
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
92 |
return f"Not Out (Missing stumps, Pitch at x: {pitch_x:.1f}, y: {pitch_y:.1f}, Impact at x: {impact_x:.1f}, y: {impact_y:.1f})", trajectory, pitch_point, impact_point
|
93 |
|
94 |
+
def generate_slow_motion(frames, trajectory, pitch_point, impact_point, output_path):
|
95 |
if not frames:
|
96 |
return None
|
97 |
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
|
98 |
out = cv2.VideoWriter(output_path, fourcc, FRAME_RATE / SLOW_MOTION_FACTOR, (frames[0].shape[1], frames[0].shape[0]))
|
99 |
|
100 |
+
for frame in frames:
|
101 |
+
# Draw full trajectory (blue dots)
|
102 |
+
if trajectory:
|
103 |
+
for x, y in trajectory:
|
104 |
+
cv2.circle(frame, (int(x), int(y)), 5, (255, 0, 0), -1)
|
105 |
|
106 |
+
# Draw pitch point (red circle with label)
|
107 |
+
if pitch_point:
|
108 |
x, y = pitch_point
|
109 |
+
cv2.circle(frame, (int(x), int(y)), 8, (0, 0, 255), -1)
|
110 |
cv2.putText(frame, "Pitch Point", (int(x) + 10, int(y) - 10),
|
111 |
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 0, 255), 2)
|
112 |
|
113 |
+
# Draw impact point (yellow circle with label)
|
114 |
+
if impact_point:
|
115 |
x, y = impact_point
|
116 |
+
cv2.circle(frame, (int(x), int(y)), 8, (0, 255, 255), -1)
|
117 |
cv2.putText(frame, "Impact Point", (int(x) + 10, int(y) + 20),
|
118 |
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 255), 2)
|
119 |
|
|
|
126 |
frames, ball_positions, debug_log = process_video(video)
|
127 |
if not frames:
|
128 |
return f"Error: Failed to process video\nDebug Log:\n{debug_log}", None
|
129 |
+
trajectory, pitch_point, impact_point, trajectory_log = estimate_trajectory(ball_positions, frames)
|
130 |
+
decision, trajectory, pitch_point, impact_point = lbw_decision(ball_positions, trajectory, frames, pitch_point, impact_point)
|
|
|
131 |
|
132 |
output_path = f"output_{uuid.uuid4()}.mp4"
|
133 |
+
slow_motion_path = generate_slow_motion(frames, trajectory, pitch_point, impact_point, output_path)
|
134 |
|
135 |
debug_output = f"{debug_log}\n{trajectory_log}"
|
136 |
return f"DRS Decision: {decision}\nDebug Log:\n{debug_output}", slow_motion_path
|
|
|
144 |
gr.Video(label="Very Slow-Motion Replay with Ball Detection (Green), Trajectory (Blue), Pitch Point (Red), Impact Point (Yellow)")
|
145 |
],
|
146 |
title="AI-Powered DRS for LBW in Local Cricket",
|
147 |
+
description="Upload a video clip of a cricket delivery to get an LBW decision and very slow-motion replay showing ball detection (green boxes), trajectory (blue dots), pitch point (red circle), and impact point (yellow circle)."
|
148 |
)
|
149 |
|
150 |
if __name__ == "__main__":
|