DRS / app.py
AjaykumarPilla's picture
Update app.py
f558e96 verified
import cv2
import numpy as np
import torch
from ultralytics import YOLO
import gradio as gr
from scipy.interpolate import interp1d
import plotly.graph_objects as go
import uuid
import os
import tempfile
# Load YOLOv8 model and resolve class index
model = YOLO("best.pt")
model.to('cuda' if torch.cuda.is_available() else 'cpu')
# Dynamically resolve ball class index
ball_class_index = None
for k, v in model.names.items():
if v.lower() == "cricketball":
ball_class_index = k
break
if ball_class_index is None:
raise ValueError("Class 'cricketBall' not found in model.names")
# Constants
STUMPS_WIDTH = 0.2286
BALL_DIAMETER = 0.073
FRAME_RATE = 20
SLOW_MOTION_FACTOR = 2
CONF_THRESHOLD = 0.2
IMPACT_ZONE_Y = 0.85
IMPACT_DELTA_Y = 50
PITCH_LENGTH = 20.12
STUMPS_HEIGHT = 0.71
MAX_POSITION_JUMP = 30
def process_video(video_path):
if not os.path.exists(video_path):
return [], [], [], "Error: Video file not found"
cap = cv2.VideoCapture(video_path)
frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
frames, ball_positions, detection_frames, debug_log = [], [], [], []
frame_count = 0
while cap.isOpened():
ret, frame = cap.read()
if not ret:
break
frame_count += 1
frames.append(frame.copy())
results = model.predict(frame, conf=CONF_THRESHOLD, imgsz=(frame_height, frame_width), iou=0.5, max_det=1)
detections = 0
for detection in results[0].boxes:
if int(detection.cls) == ball_class_index:
detections += 1
if detections == 1:
x1, y1, x2, y2 = detection.xyxy[0].cpu().numpy()
ball_positions.append([(x1 + x2) / 2, (y1 + y2) / 2])
detection_frames.append(frame_count - 1)
cv2.rectangle(frame, (int(x1), int(y1)), (int(x2), int(y2)), (0, 255, 0), 2)
frames[-1] = frame
debug_log.append(f"Frame {frame_count}: {detections} ball detections")
cap.release()
if not ball_positions:
debug_log.append("No balls detected in any frame")
else:
debug_log.append(f"Total ball detections: {len(ball_positions)}")
debug_log.append(f"Video resolution: {frame_width}x{frame_height}")
return frames, ball_positions, detection_frames, "\n".join(debug_log)
def find_bounce_point(ball_coords):
for i in range(1, len(ball_coords) - 1):
if ball_coords[i-1][1] < ball_coords[i][1] > ball_coords[i+1][1]:
return ball_coords[i]
return ball_coords[len(ball_coords)//3] # fallback
def lbw_decision(ball_positions, trajectory, frames, pitch_point, impact_point):
if not frames or not trajectory or len(ball_positions) < 2:
return "Not enough data", trajectory, pitch_point, impact_point
frame_height, frame_width = frames[0].shape[:2]
stumps_x = frame_width / 2
stumps_y = frame_height * 0.9
stumps_width_pixels = frame_width * (STUMPS_WIDTH / 3.0)
pitch_x, _ = pitch_point
impact_x, impact_y = impact_point
if pitch_x < stumps_x - stumps_width_pixels / 2 or pitch_x > stumps_x + stumps_width_pixels / 2:
return f"Not Out (Pitched outside line)", trajectory, pitch_point, impact_point
if impact_x < stumps_x - stumps_width_pixels / 2 or impact_x > stumps_x + stumps_width_pixels / 2:
return f"Not Out (Impact outside line)", trajectory, pitch_point, impact_point
for x, y in trajectory:
if abs(x - stumps_x) < stumps_width_pixels / 2 and abs(y - stumps_y) < frame_height * 0.1:
return f"Out (Ball projected to hit stumps)", trajectory, pitch_point, impact_point
return f"Not Out (Missing stumps)", trajectory, pitch_point, impact_point
def estimate_trajectory(ball_positions, detection_frames, frame_height, frame_width):
if len(ball_positions) < 2:
return None, None, None, "Error: Not enough ball detections"
filtered_positions = [ball_positions[0]]
filtered_frames = [detection_frames[0]]
for i in range(1, len(ball_positions)):
prev, curr = filtered_positions[-1], ball_positions[i]
if np.linalg.norm(np.array(curr) - np.array(prev)) <= MAX_POSITION_JUMP:
filtered_positions.append(curr)
filtered_frames.append(detection_frames[i])
if len(filtered_positions) < 2:
return None, None, None, "Error: Filtered detections too few"
x_vals = [p[0] for p in filtered_positions]
y_vals = [p[1] for p in filtered_positions]
times = np.array(filtered_frames) / FRAME_RATE
try:
fx = interp1d(times, x_vals, kind='cubic', fill_value="extrapolate")
fy = interp1d(times, y_vals, kind='cubic', fill_value="extrapolate")
except Exception as e:
return None, None, None, f"Interpolation error: {str(e)}"
total_frames = max(filtered_frames) - min(filtered_frames) + 1
t_full = np.linspace(times[0], times[-1], max(5, total_frames * SLOW_MOTION_FACTOR))
x_full = fx(t_full)
y_full = fy(t_full)
trajectory = list(zip(x_full, y_full))
pitch_point = find_bounce_point(filtered_positions)
impact_point = filtered_positions[-1]
return trajectory, pitch_point, impact_point, "Trajectory estimated successfully"
def generate_replay(frames, trajectory, pitch_point, impact_point, detection_frames):
if not frames or not trajectory:
return None
temp_file = os.path.join(tempfile.gettempdir(), f"drs_output_{uuid.uuid4()}.mp4")
height, width = frames[0].shape[:2]
out = cv2.VideoWriter(temp_file, cv2.VideoWriter_fourcc(*'mp4v'), FRAME_RATE / SLOW_MOTION_FACTOR, (width, height))
min_frame = min(detection_frames)
max_frame = max(detection_frames)
total_frames = max_frame - min_frame + 1
traj_per_frame = max(1, len(trajectory) // total_frames)
indices = [min(i * traj_per_frame, len(trajectory)-1) for i in range(total_frames)]
for i, frame in enumerate(frames):
idx = i - min_frame
if 0 <= idx < len(indices):
end_idx = indices[idx]
points = np.array(trajectory[:end_idx+1], dtype=np.int32).reshape((-1, 1, 2))
cv2.polylines(frame, [points], False, (255, 0, 0), 2)
if pitch_point and i == detection_frames[0]:
cv2.circle(frame, tuple(map(int, pitch_point)), 6, (0, 0, 255), -1)
if impact_point and i == detection_frames[-1]:
cv2.circle(frame, tuple(map(int, impact_point)), 6, (0, 255, 255), -1)
for _ in range(SLOW_MOTION_FACTOR):
out.write(frame)
out.release()
return temp_file
def drs_review(video):
frames, ball_positions, detection_frames, debug_log = process_video(video)
if not frames or not ball_positions:
return "No frames or detections found.", None
frame_height, frame_width = frames[0].shape[:2]
trajectory, pitch_point, impact_point, log = estimate_trajectory(ball_positions, detection_frames, frame_height, frame_width)
if not trajectory:
return f"{log}\n{debug_log}", None
decision, _, _, _ = lbw_decision(ball_positions, trajectory, frames, pitch_point, impact_point)
replay_path = generate_replay(frames, trajectory, pitch_point, impact_point, detection_frames)
result_log = f"DRS Decision: {decision}\n\n{log}\n\n{debug_log}"
return result_log, replay_path
# Gradio Interface
iface = gr.Interface(
fn=drs_review,
inputs=gr.Video(label="Upload Cricket Delivery Video"),
outputs=[
gr.Textbox(label="DRS Result and Debug Info"),
gr.Video(label="Replay with Trajectory & Decision")
],
title="GullyDRS - AI-Powered LBW Review",
description="Upload a cricket delivery video. The system will track the ball, estimate trajectory, and return a replay with an OUT/NOT OUT decision."
)
if __name__ == "__main__":
iface.launch()