Model 7

test_7s.py

@@ -0,0 +1,167 @@
+import cv2
+import os
+import argparse
+from pathlib import Path
+import subprocess
+import numpy as np
+import shutil
+
+def split_image(img_path, stride=32):
+    img = cv2.imread(img_path)
+    if img is None:
+        raise FileNotFoundError(f"The image at {img_path} could not be loaded. Please check the path and file.")
+    h, w, _ = img.shape
+    pad_h = (stride - h % stride) % stride
+    pad_w = (stride - w % stride) % stride
+    padded_img = cv2.copyMakeBorder(img, 0, pad_h, 0, pad_w, cv2.BORDER_CONSTANT, value=(0,0,0))
+    h, w = padded_img.shape[:2]
+    return [padded_img[:h//2, :w//2], padded_img[:h//2, w//2:], padded_img[h//2:, :w//2], padded_img[h//2:, w//2:]]
+
+def save_image_parts(img_parts, temp_dir):
+    part_paths = []
+    for idx, part in enumerate(img_parts):
+        part_path = os.path.join(temp_dir, f"part_{idx + 1}.jpg")
+        cv2.imwrite(part_path, part)
+        part_paths.append(part_path)
+    return part_paths
+
+def parse_predictions(pred_path):
+    predictions = []
+    with open(pred_path, 'r') as file:
+        for line in file:
+            data = line.strip().split()
+            if len(data) == 6:
+                predictions.append([int(data[0])] + list(map(float, data[1:])))
+            elif len(data) == 5:  # Handle missing confidence
+                predictions.append([int(data[0])] + list(map(float, data[1:])) + [1.0])  # Default confidence to 1.0
+    return predictions
+
+def draw_predictions_on_original(image, predictions, part_idx, img_shape, part_shape):
+    h_half, w_half = img_shape[0] // 2, img_shape[1] // 2
+    x_offset = (part_idx % 2) * w_half
+    y_offset = (part_idx // 2) * h_half
+
+    bboxes = []
+    for pred in predictions:
+        class_id, x_center, y_center, width, height, conf = pred
+        # Transform normalized coordinates to absolute coordinates
+        abs_x_center = x_center * part_shape[1] + x_offset
+        abs_y_center = y_center * part_shape[0] + y_offset
+        abs_width = width * part_shape[1]
+        abs_height = height * part_shape[0]
+
+        top_left = (int(abs_x_center - abs_width / 2), int(abs_y_center - abs_height / 2))
+        bottom_right = (int(abs_x_center + abs_width / 2), int(abs_y_center + abs_height / 2))
+
+        # Set bounding box color and label based on class_id
+        if class_id == 0:
+            color = (0, 255, 0)  # Green for Healthy
+            label = "Healthy"
+        else:
+            color = (0, 0, 255)  # Red for Sick
+            label = "Sick"
+
+        cv2.rectangle(image, top_left, bottom_right, color, 2)
+        cv2.putText(image, f"{label} {conf:.2f}", (top_left[0], top_left[1] - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)
+
+        # Normalize coordinates relative to the full image
+        norm_x_center = abs_x_center / img_shape[1]
+        norm_y_center = abs_y_center / img_shape[0]
+        norm_width = abs_width / img_shape[1]
+        norm_height = abs_height / img_shape[0]
+
+        # Collect bounding box information
+        bboxes.append((class_id, norm_x_center, norm_y_center, norm_width, norm_height, conf))
+
+    return bboxes
+
+def process_image_parts(part_paths, full_img_path, args):
+    full_img = cv2.imread(full_img_path)
+    img_shape = full_img.shape
+    part_shape = (img_shape[0] // 2, img_shape[1] // 2)
+
+    all_bboxes = []
+    for idx, part_path in enumerate(part_paths):
+        # Prepare the command line arguments list for subprocess.run()
+        command = [
+            'python', '/content/drive/MyDrive/development/models/model-7/yolov7/run.py',
+            '--source', part_path,
+            '--img-size', str(args.img_size),
+            '--save-txt',  # Ensure predictions are saved to txt files
+            '--save-conf',  # Ensure confidence scores are saved
+            '--project', args.temp_dir,  # Output directory
+            '--name', 'outputs',  # Subdirectory for results
+            '--exist-ok'  # Prevent creating new directories each time
+        ]
+
+        # If weights is a list, extend the command list appropriately
+        if isinstance(args.weights, list):
+            command.extend(['--weights'] + args.weights)
+        else:
+            command.extend(['--weights', args.weights])
+
+        # Execute the model script via subprocess
+        subprocess.run(command, check=True, text=True)
+
+        # Construct the prediction file path
+        pred_path = os.path.join(args.temp_dir, 'outputs', 'labels', os.path.basename(part_path).replace('.jpg', '.txt'))
+
+        predictions = parse_predictions(pred_path) if os.path.exists(pred_path) else []
+        bboxes = draw_predictions_on_original(full_img, predictions, idx, img_shape, part_shape)
+        all_bboxes.extend(bboxes)
+
+    result_path = os.path.join(args.project, args.name, os.path.basename(full_img_path))
+    cv2.imwrite(result_path, full_img)
+    print(f"Combined image with predictions saved to: {result_path}")
+
+    # Ensure the labels directory exists
+    labels_dir = os.path.join(args.project, args.name, 'labels')
+    Path(labels_dir).mkdir(parents=True, exist_ok=True)
+
+    # Save bounding boxes to a text file in the labels directory
+    bbox_txt_path = os.path.join(labels_dir, os.path.basename(full_img_path).replace('.jpg', '.txt'))
+    with open(bbox_txt_path, 'w') as f:
+        for bbox in all_bboxes:
+            f.write(' '.join(map(str, bbox)) + '\n')
+    print(f"Bounding box values saved to: {bbox_txt_path}")
+
+def clear_labels_directory(temp_dir):
+    labels_dir = os.path.join(temp_dir, 'outputs', 'labels')
+    if os.path.exists(labels_dir):
+        shutil.rmtree(labels_dir)
+    Path(labels_dir).mkdir(parents=True, exist_ok=True)
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--source', type=str, required=True, help='Source image path or directory')
+    parser.add_argument('--weights', nargs='+', type=str, default=['yolov7.pt'], help='Path(s) to model weights')
+    parser.add_argument('--img-size', type=int, default=640, help='Inference size (pixels)')
+    parser.add_argument('--save-txt', action='store_true', help='Save results to .txt file')
+    parser.add_argument('--save-conf', action='store_true', help='Save confidence scores')
+    parser.add_argument('--project', type=str, required=True, help='Project directory')
+    parser.add_argument('--name', type=str, required=True, help='Run name')
+    parser.add_argument('--temp-dir', type=str, default='temp_parts', help='Temporary directory for image parts')
+    args = parser.parse_args()
+
+    # Ensure necessary directories exist
+    Path(args.temp_dir).mkdir(parents=True, exist_ok=True)
+    Path(os.path.join(args.project, args.name)).mkdir(parents=True, exist_ok=True)
+
+    if os.path.isdir(args.source):
+        for filename in os.listdir(args.source):
+            file_path = os.path.join(args.source, filename)
+            if os.path.isfile(file_path) and file_path.lower().endswith(('.png', '.jpg', '.jpeg')):
+                print(f"Processing image: {file_path}")
+                clear_labels_directory(args.temp_dir)
+                img_parts = split_image(file_path)
+                part_paths = save_image_parts(img_parts, args.temp_dir)
+                process_image_parts(part_paths, file_path, args)
+    else:
+        img = cv2.imread(args.source)
+        clear_labels_directory(args.temp_dir)
+        img_parts = split_image(args.source)
+        part_paths = save_image_parts(img_parts, args.temp_dir)
+        process_image_parts(part_paths, args.source, args)
+
+if __name__ == '__main__':
+    main()

train.py

@@ -1,3 +1,4 @@
+
 import argparse
 import logging
 import math

utils/google_utils.py

@@ -19,7 +19,8 @@ def gsutil_getsize(url=''):
 def attempt_download(file, repo='WongKinYiu/yolov7'):
     # Attempt file download if does not exist
     file = Path(str(file).strip().replace("'", '').lower())
-
+    if os.path.exists(file):
+        return  # File exists, no need to download
     if not file.exists():
         try:
             response = requests.get(f'https://api.github.com/repos/{repo}/releases/latest').json()  # github api
@@ -28,7 +29,10 @@ def attempt_download(file, repo='WongKinYiu/yolov7'):
         except:  # fallback plan
             assets = ['yolov7.pt', 'yolov7-tiny.pt', 'yolov7x.pt', 'yolov7-d6.pt', 'yolov7-e6.pt', 
                       'yolov7-e6e.pt', 'yolov7-w6.pt']
-            tag = subprocess.check_output('git tag', shell=True).decode().split()[-1]
+            try:
+                tag = subprocess.check_output('git tag', shell=True).decode().split()[-1]
+            except subprocess.CalledProcessError:
+                tag = 'v0.0'  # default tag if git command fails
 
         name = file.name
         if name in assets: