Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -3,32 +3,26 @@ import torch
|
|
3 |
import gradio as gr
|
4 |
import numpy as np
|
5 |
from ultralytics import YOLO
|
|
|
6 |
|
7 |
# Load YOLOv8 model
|
8 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
9 |
model = YOLO('./data/best.pt') # Path to your model
|
10 |
model.to(device)
|
11 |
|
12 |
-
#
|
13 |
frames_with_detections = []
|
14 |
-
detection_counts = []
|
15 |
|
16 |
-
# Define the function
|
17 |
def process_video(video):
|
18 |
-
#
|
19 |
-
input_video = cv2.VideoCapture(video)
|
20 |
-
|
21 |
-
# Get frame width, height, and fps from input video
|
22 |
frame_width = int(input_video.get(cv2.CAP_PROP_FRAME_WIDTH))
|
23 |
frame_height = int(input_video.get(cv2.CAP_PROP_FRAME_HEIGHT))
|
24 |
fps = input_video.get(cv2.CAP_PROP_FPS)
|
25 |
|
26 |
-
# Resize to reduce
|
27 |
-
new_width, new_height = 640, 480
|
28 |
-
frame_width, frame_height = new_width, new_height
|
29 |
-
|
30 |
-
# Track detected objects by their bounding box coordinates
|
31 |
-
detected_boxes = set()
|
32 |
|
33 |
while True:
|
34 |
# Read a frame from the video
|
@@ -36,56 +30,54 @@ def process_video(video):
|
|
36 |
if not ret:
|
37 |
break # End of video
|
38 |
|
39 |
-
# Resize the frame
|
40 |
frame = cv2.resize(frame, (new_width, new_height))
|
41 |
|
42 |
# Perform inference on the frame
|
43 |
results = model(frame) # Automatically uses GPU if available
|
44 |
|
45 |
-
#
|
46 |
-
if len(results[0].boxes) > 0:
|
47 |
-
# Get the bounding boxes
|
48 |
-
boxes = results[0].boxes.xyxy.cpu().numpy() # Get xyxy coordinates
|
49 |
|
50 |
-
#
|
51 |
-
|
52 |
-
x1, y1, x2, y2 = box
|
53 |
-
detection_box = (x1, y1, x2, y2)
|
54 |
|
55 |
-
|
56 |
-
|
57 |
-
# Add the box to the set to avoid repeating the detection
|
58 |
-
detected_boxes.add(detection_box)
|
59 |
|
60 |
-
|
61 |
-
|
62 |
|
63 |
-
|
64 |
-
|
|
|
|
|
|
|
|
|
65 |
|
66 |
-
|
67 |
-
|
|
|
68 |
|
69 |
-
|
70 |
-
|
|
|
|
|
|
|
|
|
71 |
|
72 |
# Release resources
|
73 |
input_video.release()
|
74 |
|
75 |
-
#
|
76 |
with gr.Blocks() as demo:
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
# Define the function to update frames in the album
|
84 |
-
def update_gallery(video):
|
85 |
-
return process_video(video) # Return frames one by one as they are detected
|
86 |
-
|
87 |
-
# Connect the video input to the gallery update
|
88 |
-
video_input.change(update_gallery, inputs=video_input, outputs=gallery_output)
|
89 |
|
90 |
# Launch the interface
|
91 |
demo.launch()
|
|
|
3 |
import gradio as gr
|
4 |
import numpy as np
|
5 |
from ultralytics import YOLO
|
6 |
+
import matplotlib.pyplot as plt
|
7 |
|
8 |
# Load YOLOv8 model
|
9 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
10 |
model = YOLO('./data/best.pt') # Path to your model
|
11 |
model.to(device)
|
12 |
|
13 |
+
# List to store frames with detections
|
14 |
frames_with_detections = []
|
|
|
15 |
|
16 |
+
# Define the function to process the video
|
17 |
def process_video(video):
|
18 |
+
# Open the video file
|
19 |
+
input_video = cv2.VideoCapture(video)
|
|
|
|
|
20 |
frame_width = int(input_video.get(cv2.CAP_PROP_FRAME_WIDTH))
|
21 |
frame_height = int(input_video.get(cv2.CAP_PROP_FRAME_HEIGHT))
|
22 |
fps = input_video.get(cv2.CAP_PROP_FPS)
|
23 |
|
24 |
+
# Resize frames to 640x480 (optional, to reduce computational load)
|
25 |
+
new_width, new_height = 640, 480
|
|
|
|
|
|
|
|
|
26 |
|
27 |
while True:
|
28 |
# Read a frame from the video
|
|
|
30 |
if not ret:
|
31 |
break # End of video
|
32 |
|
33 |
+
# Resize the frame
|
34 |
frame = cv2.resize(frame, (new_width, new_height))
|
35 |
|
36 |
# Perform inference on the frame
|
37 |
results = model(frame) # Automatically uses GPU if available
|
38 |
|
39 |
+
# If there are detections
|
40 |
+
if len(results[0].boxes) > 0:
|
41 |
+
boxes = results[0].boxes.xyxy.cpu().numpy() # Get the bounding boxes
|
|
|
42 |
|
43 |
+
# Annotate the frame with bounding boxes
|
44 |
+
annotated_frame = results[0].plot()
|
|
|
|
|
45 |
|
46 |
+
# Convert the frame to RGB
|
47 |
+
annotated_frame_rgb = cv2.cvtColor(annotated_frame, cv2.COLOR_BGR2RGB)
|
|
|
|
|
48 |
|
49 |
+
# Append the frame with detection to list
|
50 |
+
frames_with_detections.append(annotated_frame_rgb)
|
51 |
|
52 |
+
# Create a simple bar chart to show the count of detected objects
|
53 |
+
fig, ax = plt.subplots()
|
54 |
+
ax.bar([1], [len(boxes)], color='blue') # Bar for the current frame detection
|
55 |
+
ax.set_xlabel('Frame')
|
56 |
+
ax.set_ylabel('Number of Detections')
|
57 |
+
ax.set_title('Detection Count per Frame')
|
58 |
|
59 |
+
# Convert plot to an image to return it in Gradio output
|
60 |
+
plt.tight_layout()
|
61 |
+
plt.close(fig)
|
62 |
|
63 |
+
# Save the plot as an image in memory
|
64 |
+
buf = np.frombuffer(fig.canvas.print_to_buffer()[0], dtype=np.uint8)
|
65 |
+
img = cv2.imdecode(buf, cv2.IMREAD_COLOR)
|
66 |
+
|
67 |
+
# Yield the detected frame and the graph at the same time
|
68 |
+
yield annotated_frame_rgb, img
|
69 |
|
70 |
# Release resources
|
71 |
input_video.release()
|
72 |
|
73 |
+
# Gradio interface
|
74 |
with gr.Blocks() as demo:
|
75 |
+
with gr.Row():
|
76 |
+
video_input = gr.Video(label="Upload Video")
|
77 |
+
gallery_output = gr.Gallery(label="Detection Album").style(columns=3) # Display images in a row
|
78 |
+
graph_output = gr.Image(label="Detection Counts Graph", type="numpy") # For displaying graph
|
79 |
+
|
80 |
+
video_input.change(process_video, inputs=video_input, outputs=[gallery_output, graph_output])
|
|
|
|
|
|
|
|
|
|
|
|
|
81 |
|
82 |
# Launch the interface
|
83 |
demo.launch()
|