jolie
Browse files- inference.py +69 -0
inference.py
ADDED
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from ultralytics import YOLO
|
2 |
+
import matplotlib.pyplot as plt
|
3 |
+
import glob
|
4 |
+
import os
|
5 |
+
|
6 |
+
def visualize_predictions(result_dir):
|
7 |
+
"""Visualize up to four prediction results."""
|
8 |
+
image_paths = glob.glob(os.path.join(result_dir, '*.jpg'))
|
9 |
+
num_images = min(4, len(image_paths))
|
10 |
+
|
11 |
+
if num_images == 0:
|
12 |
+
print("No images found for visualization.")
|
13 |
+
return
|
14 |
+
|
15 |
+
plt.figure(figsize=(15, 12))
|
16 |
+
for i, image_path in enumerate(image_paths[:num_images]):
|
17 |
+
image = plt.imread(image_path)
|
18 |
+
plt.subplot(2, 2, i + 1)
|
19 |
+
plt.imshow(image)
|
20 |
+
plt.axis('off')
|
21 |
+
plt.tight_layout()
|
22 |
+
plt.show()
|
23 |
+
|
24 |
+
def run_inference(checkpoint_path, inference_source='combined_dataset/images/valid', inference_name='yolo_infer_last'):
|
25 |
+
"""Run inference using the saved checkpoint."""
|
26 |
+
if not os.path.exists(checkpoint_path):
|
27 |
+
print(f"Checkpoint '{checkpoint_path}' does not exist. Please ensure the path is correct.")
|
28 |
+
return
|
29 |
+
|
30 |
+
print(f"Loading the model from '{checkpoint_path}'...")
|
31 |
+
try:
|
32 |
+
# Load the model with the saved weights
|
33 |
+
model = YOLO(checkpoint_path)
|
34 |
+
print("Model loaded successfully.")
|
35 |
+
except Exception as e:
|
36 |
+
print(f"Error loading model: {e}")
|
37 |
+
return
|
38 |
+
|
39 |
+
# Verify inference source
|
40 |
+
if not os.path.exists(inference_source):
|
41 |
+
print(f"Inference source '{inference_source}' does not exist. Please provide a valid path.")
|
42 |
+
return
|
43 |
+
|
44 |
+
print(f"Running inference on '{inference_source}'...")
|
45 |
+
try:
|
46 |
+
results = model.predict(
|
47 |
+
source=inference_source,
|
48 |
+
save=True,
|
49 |
+
project='runs/predict',
|
50 |
+
name=inference_name,
|
51 |
+
exist_ok=True
|
52 |
+
)
|
53 |
+
print("Inference completed.")
|
54 |
+
except Exception as e:
|
55 |
+
print(f"Error during inference: {e}")
|
56 |
+
return
|
57 |
+
|
58 |
+
# Visualize predictions
|
59 |
+
visualize_predictions(os.path.join('runs', 'predict', inference_name))
|
60 |
+
|
61 |
+
def main():
|
62 |
+
# Define the path to the checkpoint
|
63 |
+
checkpoint_path = 'Edutech/train/weights/last.pt' # Adjust the path if necessary
|
64 |
+
|
65 |
+
# Run inference
|
66 |
+
run_inference(checkpoint_path, inference_name='yolo_infer_last')
|
67 |
+
|
68 |
+
if __name__ == "__main__":
|
69 |
+
main()
|