Updates

rpi/board-detector/main.py

@@ -1,5 +1,6 @@
 #!/usr/bin/env python3
-from paths import *
+import os
+import random
 
 from ultralytics import YOLO
 import cv2
@@ -62,18 +63,31 @@ def prediction_to_fen(results, width, height):
 
 
 if __name__ == "__main__":
+    model_path = "../assets/models/unified-nano-refined.pt"
+    img_folder = "../training/datasets/pieces/unified/test/images/"
+    save_folder = "./results"
+    os.makedirs(save_folder, exist_ok=True)
 
-    img = cv2.imread(img_path)
-    height, width = img.shape[:2]
+    test_images = os.listdir(img_folder)
 
-    model = YOLO(model_path)
-    results = model.predict(source=img_path, conf=0.5)
+    for i in range(0, 10):
+        rnd = random.randint(0, len(test_images) - 1)
+        img_path = os.path.join(img_folder, test_images[rnd])
+        save_path = os.path.join(save_folder, test_images[rnd])
 
-    #fen = prediction_to_fen(results, height, width)
-    #print("Predicted FEN:", fen)
+        img = cv2.imread(img_path)
+        height, width = img.shape[:2]
 
-    annotated_image = results[0].plot()  # Annotated image as NumPy array
-    cv2.namedWindow("YOLO Predictions", cv2.WINDOW_NORMAL)  # make window resizable
-    cv2.imshow("YOLO Predictions", annotated_image)
-    cv2.waitKey(0)
-    cv2.destroyAllWindows()
+        model = YOLO(model_path)
+        results = model.predict(source=img_path, conf=0.5)
+
+        #fen = prediction_to_fen(results, height, width)
+        #print("Predicted FEN:", fen)
+
+        annotated_image = results[0].plot()
+        cv2.imwrite(save_path, annotated_image)
+        #cv2.namedWindow("YOLO Predictions", cv2.WINDOW_NORMAL)
+        #cv2.imshow("YOLO Predictions", annotated_image)
+
+        cv2.waitKey(0)
+        cv2.destroyAllWindows()

rpi/board-detector/paths.py

@@ -1,3 +0,0 @@
-model_path = "C:/Users/Laurent/Desktop/board-mate/rpi/assets/models/epoch-200.pt"
-#img_path = "./test/4.jpg"
-img_path = "../training/datasets/unified/train/images/WIN_20221220_11_27_27_Pro_jpg.rf.4f01cb68c8944ef1c4c7dc57847b4cd3.jpg"

rpi/board-detector/realtime_detect.py

@@ -1,11 +1,14 @@
 from ultralytics import YOLO
-from paths import *  # make sure model_path is defined here
 import cv2
 
 if __name__ == "__main__":
 
+    corner_model_path = "../assets/models/corner.pt"
+    pieces_model_path = "../assets/models/unified-nano-refined.pt"
+
     print("Initializing model...")
-    model = YOLO(model_path)
+    corner_model = YOLO(corner_model_path)
+    pieces_model = YOLO(pieces_model_path)
 
     print("Initializing camera...")
     cap = cv2.VideoCapture(0)
@@ -29,11 +32,13 @@ if __name__ == "__main__":
         # Optional: resize frame to improve YOLO performance
         # frame = cv2.resize(frame, (416, 416))
 
-        results = model.predict(source=frame, conf=0.5)
+        corner_result = corner_model.predict(source=frame, conf=0.6)
+        pieces_result = pieces_model.predict(source=frame, conf=0.6)
 
-        annotated_frame = results[0].plot()  # annotated frame as NumPy array
+        corner_annotated_frame = corner_result[0].plot()
+        pieces_annotated_frame = pieces_result[0].plot(img=corner_annotated_frame)
 
-        cv2.imshow("Predictions", annotated_frame)
+        cv2.imshow("Predictions", pieces_annotated_frame)
         cv2.resizeWindow("Predictions", 640, 640)
         if cv2.waitKey(1) & 0xFF == ord('q'):
             break

rpi/requirements.txt

@@ -13,3 +13,5 @@ pyserial
 opencv-python
 numpy
 ultralytics
+
+torch

rpi/training/chess.yaml

@@ -1,6 +0,0 @@
-train: C:/Users/Laurent/Desktop/board-mate/rpi/training/datasets/roboflow/labels-bck-bck-bck
-val: C:/Users/Laurent/Desktop/board-mate/rpi/training/datasets/roboflow/labels-bck-bck-bck
-
-nc: 12
-names: ['w_pawn','w_knight','w_bishop','w_rook','w_queen','w_king',
-        'b_pawn','b_knight','b_bishop','b_rook','b_queen','b_king']

rpi/training/datasets/chesscog/data.yaml

@@ -1,7 +0,0 @@
-train: ../train/
-val: ../valid/
-test: ../test/
-
-nc: 12
-names: ['w_pawn','w_knight','w_bishop','w_rook','w_queen','w_king',
-        'b_pawn','b_knight','b_bishop','b_rook','b_queen','b_king']

rpi/training/datasets/unified/data.yaml

@@ -1,7 +0,0 @@
-train: ../train/
-val: ../valid/
-test: ../test/
-
-nc: 12
-names: ['w_pawn','w_knight','w_bishop','w_rook','w_queen','w_king',
-        'b_pawn','b_knight','b_bishop','b_rook','b_queen','b_king']

rpi/training/labelizer.py

@@ -1,70 +0,0 @@
-import os
-import cv2
-from ultralytics import YOLO
-
-# --------------------------
-# Configuration
-# --------------------------
-model_path = "models/bck/best-3.pt"  # your trained YOLO model
-images_dir = "C:/Users/Laurent/Desktop/board-mate/rpi/training/datasets/universe/train/images"
-labels_dir = "C:/Users/Laurent/Desktop/board-mate/rpi/training/datasets/universe/train/labels"
-img_width = 640
-img_height = 640
-
-os.makedirs(labels_dir, exist_ok=True)
-
-# --------------------------
-# Load model
-# --------------------------
-model = YOLO(model_path)
-
-# --------------------------
-# Mapping YOLO class index -> piece name (optional)
-# --------------------------
-names = ['w_pawn','w_knight','w_bishop','w_rook','w_queen','w_king',
-         'b_pawn','b_knight','b_bishop','b_rook','b_queen','b_king']
-
-# --------------------------
-# Process images
-# --------------------------
-for img_file in os.listdir(images_dir):
-    if not img_file.lower().endswith((".png", ".jpg", ".jpeg")):
-        continue
-
-    img_path = os.path.join(images_dir, img_file)
-    img = cv2.imread(img_path)
-    if img is None:
-        print(f"Failed to read {img_file}")
-        continue
-
-    height, width = img.shape[:2]
-
-    # Run YOLO detection
-    results = model(img)
-    res = results[0]
-
-    lines = []
-    boxes = res.boxes.xyxy.cpu().numpy()  # [x1, y1, x2, y2]
-    classes = res.boxes.cls.cpu().numpy()
-    confs = res.boxes.conf.cpu().numpy()
-
-    for box, cls, conf in zip(boxes, classes, confs):
-        if conf < 0.5:  # skip low-confidence predictions
-            continue
-
-        x1, y1, x2, y2 = box
-        x_center = (x1 + x2) / 2 / width
-        y_center = (y1 + y2) / 2 / height
-        w_norm = (x2 - x1) / width
-        h_norm = (y2 - y1) / height
-
-        lines.append(f"{int(cls)} {x_center:.6f} {y_center:.6f} {w_norm:.6f} {h_norm:.6f}")
-
-    # Save YOLO .txt file with same basename as image
-    txt_path = os.path.join(labels_dir, os.path.splitext(img_file)[0] + ".txt")
-    with open(txt_path, "w") as f:
-        f.write("\n".join(lines))
-
-    print(f"Pre-labeled {img_file} -> {txt_path}")
-
-print("All images have been pre-labeled!")

rpi/training/move_image.py

@@ -1,25 +0,0 @@
-import os
-import shutil
-
-# ----------------------------
-# Configuration
-# ----------------------------
-
-source_folder = "datasets/twhpv/valid/images"
-destination_folder = "datasets/_unified/valid/images"
-
-os.makedirs(destination_folder, exist_ok=True)
-
-# Supported image extensions
-image_extensions = [".jpg", ".jpeg", ".png", ".bmp", ".gif"]
-
-# ----------------------------
-# Copy images
-# ----------------------------
-for filename in os.listdir(source_folder):
-    if any(filename.lower().endswith(ext) for ext in image_extensions):
-        src_path = os.path.join(source_folder, filename)
-        dst_path = os.path.join(destination_folder, filename)
-        shutil.copy2(src_path, dst_path)  # copy2 preserves metadata
-
-print(f"All images copied to '{destination_folder}'")

rpi/training/sort_labels.py

@@ -1,59 +0,0 @@
-import os
-
-# ----------------------------
-# Configuration
-# ----------------------------
-
-src_dir = "datasets/visiope/test/labels"
-dest_dir = "datasets/_unified/test/labels"
-
-os.makedirs(dest_dir, exist_ok=True)
-
-# Reference class order you want to follow
-"""[
-'w_pawn','w_knight','w_bishop','w_rook','w_queen','w_king',
-'b_pawn','b_knight','b_bishop','b_rook','b_queen','b_king'
-]"""
-
-reference_classes = [
-    'w_pawn','w_knight','w_bishop','w_rook','w_queen','w_king',
-    'b_pawn','b_knight','b_bishop','b_rook','b_queen','b_king'
-]
-
-# Current class order in your dataset (change this to match your dataset!)
-current_classes = ['bishop', 'black-bishop', 'black-king', 'black-knight', 'black-pawn', 'black-queen', 'black-rook',
-        'white-bishop', 'white-king', 'white-knight', 'white-pawn', 'white-queen', 'white-rook']
-
-# ----------------------------
-# Build index mapping
-# ----------------------------
-index_map = {current_classes.index(cls): reference_classes.index(cls) for cls in current_classes}
-
-# ----------------------------
-# Process each label file
-# ----------------------------
-
-count = 0
-for filename in os.listdir(src_dir):
-    if filename.endswith(".txt"):
-        input_path = os.path.join(src_dir, filename)
-        output_path = os.path.join(dest_dir, filename)
-
-        with open(input_path, "r") as f:
-            lines = f.readlines()
-
-        new_lines = []
-        for line in lines:
-            parts = line.strip().split()
-            old_idx = int(parts[0])
-            new_idx = index_map[old_idx]
-            new_lines.append(" ".join([str(new_idx)] + parts[1:]))
-
-        with open(output_path, "w") as f:
-            f.write("\n".join(new_lines))
-
-        if count%100 == 0:
-            print(count)
-        count += 1
-
-print(f"All labels remapped and saved to '{dest_dir}'")

rpi/training/training.py

@@ -1,18 +0,0 @@
-from ultralytics import YOLO
-
-def main():
-    model = YOLO("models/yolo11n.pt")
-    model.train(
-        data="./datasets/unified/data.yaml",
-        epochs=200,
-        patience=30,
-        imgsz=640,
-        batch=12,
-        device=0,
-        project="result",
-        name="unified-training",
-        exist_ok=True
-    )
-
-if __name__ == "__main__":
-    main()

rpi/training/training_v11_n.py

@@ -0,0 +1,19 @@
+from ultralytics import YOLO
+
+def main():
+    model = YOLO("models/unified-nano.pt")
+    model.train(
+        data="./datasets/pieces/unified/data.yaml",
+        epochs=150,
+        patience=20,
+        imgsz=640,
+        batch=18,
+        save_period=10,
+        project="result",
+        name="unified-nano-refined",
+        exist_ok=True,
+        device = 0
+    )
+
+if __name__ == "__main__":
+    main()

rpi/training/training_v11_s.py

@@ -0,0 +1,31 @@
+from ultralytics import YOLO
+
+def main():
+    model = YOLO("models/yolo11s.pt")
+    model.train(
+        data="./datasets/pieces/unified/data.yaml",
+        epochs=150,
+        patience=20,
+        imgsz=640,
+        batch=12,
+        save_period=10,
+        project="result",
+        name="unified-small",
+        exist_ok=True,
+        device=0,
+        augment=True,
+        flipud=0.5,          # vertical flip with 50% probability
+        fliplr=0.5,          # horizontal flip with 50% probability
+        hsv_h=0.015,         # change hue
+        hsv_s=0.7,           # change saturation
+        hsv_v=0.4,           # change brightness/value
+        degrees=10.0,        # random rotation ±10 degrees
+        translate=0.1,       # translation ±10%
+        scale=0.1,           # scaling ±10%
+        shear=2.0,           # shear ±2 degrees
+        mosaic=1.0,          # mosaic augmentation
+        mixup=0.5            # mixup augmentation
+    )
+
+if __name__ == "__main__":
+    main()

rpi/training/utils/debug.py

@@ -0,0 +1,26 @@
+from pathlib import Path
+
+label_dir = Path("./datasets/your_dataset/val/labels")
+
+def valid_line(line):
+    parts = line.split()
+    if len(parts) != 5:
+        return False
+    cls, x, y, w, h = map(float, parts)
+    return 0 <= cls and 0 < w <= 1 and 0 < h <= 1
+
+
+empty = []
+for p in label_dir.glob("*.txt"):
+    if p.stat().st_size == 0:
+        empty.append(p.name)
+
+print(f"Empty label files: {len(empty)}")
+
+invalid = []
+for p in label_dir.glob("*.txt"):
+    lines = p.read_text().strip().splitlines()
+    if not any(valid_line(l) for l in lines):
+        invalid.append(p.name)
+
+print(f"Effectively empty labels: {len(invalid)}")

rpi/training/utils/remove_labels.py

@@ -0,0 +1,33 @@
+import os
+
+labels_dir = "../datasets/corners/Outer Chess Corners.v1i.yolov11/valid/labels"
+label_to_be_removed = 1
+
+for filename in os.listdir(labels_dir):
+    if not filename.endswith(".txt"):
+        continue
+
+    txt_path = os.path.join(labels_dir, filename)
+    new_lines = []
+
+    with open(txt_path, "r") as f:
+        lines = f.readlines()
+
+    for line in lines:
+        parts = line.strip().split()
+
+        if len(parts) < 5:
+            continue
+
+        cls = int(parts[0])
+        if cls == label_to_be_removed:
+            print(f"{parts} found in {filename}")
+            continue
+
+        new_lines.append(" ".join([str(cls)] + parts[1:]))
+
+    # Overwrite file with updated indices
+    with open(txt_path, "w") as f:
+        f.write("\n".join(new_lines))
+
+print("All label files have been adjusted!")

rpi/training/utils/remover.py

@@ -0,0 +1,16 @@
+import os
+
+# Paths to the folders
+folder_to_check = "./datasets/roboflow/train/labels"
+folder_with_files = "./datasets/unified/train/labels"
+
+files_to_check = set(os.listdir(folder_to_check))
+
+for filename in os.listdir(folder_with_files):
+    file_path = os.path.join(folder_with_files, filename)
+    if filename in files_to_check and os.path.isfile(file_path):
+        try:
+            os.remove(file_path)
+            print(f"Deleted: {file_path}")
+        except Exception as e:
+            print(f"Error deleting {file_path}: {e}")

rpi/training/utils/sort_labels.py

@@ -0,0 +1,71 @@
+import os
+import shutil
+
+
+def copy_images(src, dest):
+    image_extensions = [".jpg", ".jpeg", ".png", ".bmp", ".gif"]
+
+    for filename in os.listdir(src):
+        if any(filename.lower().endswith(ext) for ext in image_extensions):
+            src_path = os.path.join(src, filename)
+            dst_path = os.path.join(dest, filename)
+            shutil.copy2(src_path, dst_path)
+
+
+def remap_labels(src, dest):
+    count = 0
+    for filename in os.listdir(src):
+        if filename.endswith(".txt"):
+            input_path = os.path.join(src, filename)
+            output_path = os.path.join(dest, filename)
+
+            with open(input_path, "r") as f:
+                lines = f.readlines()
+
+            new_lines = []
+            for line in lines:
+                parts = line.strip().split()
+                old_idx = int(parts[0])
+                new_idx = index_map[old_idx]
+                new_lines.append(" ".join([str(new_idx)] + parts[1:]))
+
+            with open(output_path, "w") as f:
+                f.write("\n".join(new_lines))
+
+            if count%100 == 0:
+                print(count)
+            count += 1
+
+        print(f"All labels remapped and saved to '{dest}'")
+
+
+if __name__ == "__main__":
+
+    src_dir = "../datasets/pieces/visualizan/"
+    dest_dir = "../datasets/pieces/unified/"
+
+    reference_classes = [
+        'w_pawn', 'w_knight', 'w_bishop', 'w_rook', 'w_queen', 'w_king',
+        'b_pawn', 'b_knight', 'b_bishop', 'b_rook', 'b_queen', 'b_king'
+    ]
+
+    current_classes = ['b_bishop', 'b_king', 'b_knight', 'b_pawn', 'b_queen', 'b_rook',
+        'w_bishop', 'w_king', 'w_knight', 'w_pawn', 'w_queen', 'w_rook']
+
+    index_map = {current_classes.index(cls): reference_classes.index(cls) for cls in current_classes}
+
+    sub_elements = os.listdir(src_dir)
+    for sub in sub_elements:
+        src_full_path = os.path.normpath(os.path.join(src_dir, sub))
+        dest_full_path = os.path.normpath(os.path.join(dest_dir, sub))
+
+        if not os.path.isdir(src_full_path): continue
+
+        src_image_folder = os.path.normpath(os.path.join(src_full_path, "images"))
+        src_labels_folder = os.path.normpath(os.path.join(src_full_path, "labels"))
+
+        dst_image_folder = os.path.normpath(os.path.join(dest_full_path, "images"))
+        dst_labels_folder = os.path.normpath(os.path.join(dest_full_path, "labels"))
+
+        copy_images(src_image_folder, dst_image_folder)
+        remap_labels(src_labels_folder, dst_labels_folder)

rpi/training/utils/verifiy.py

@@ -0,0 +1,47 @@
+import os
+
+if __name__ == "__main__":
+    trg_dir = "../datasets/pieces/unified/train/labels"
+    src_dir = "../datasets/pieces/khalid/train/labels"
+
+    trg_labels = [
+        'w_pawn','w_knight','w_bishop','w_rook','w_queen','w_king',
+        'b_pawn','b_knight','b_bishop','b_rook','b_queen','b_king'
+    ]
+
+    src_labels = [
+        'b_bishop', 'b_king', 'b_knight', 'b_queen', 'b_rook', 'b_pawn',
+        'w_bishop', 'w_king', 'w_knight', 'w_queen', 'w_rook', 'w_pawn'
+    ]
+
+    trg_files = os.listdir(trg_dir)
+    src_files = os.listdir(src_dir)
+
+    for src_file in src_files:
+        trg_file = os.path.abspath(os.path.join(trg_dir, src_file))
+        src_file = os.path.abspath(os.path.join(src_dir, src_file))
+
+        trg_lines = []
+        src_lines = []
+
+        with open(src_file, "r") as f:
+            src_lines = f.readlines().copy()
+
+        with open(trg_file, "r") as f:
+            trg_lines = f.readlines().copy()
+
+        for i in range(0, len(trg_lines)):
+            trg_line = trg_lines[i]
+            src_line  = src_lines[i]
+
+            trg_label_index = int(trg_line.strip().split(" ")[0])
+            src_label_index = int(src_line.strip().split(" ")[0])
+
+            trg_label_value = trg_labels[trg_label_index]
+            src_label_value = src_labels[src_label_index]
+
+            if trg_label_value != src_label_value :
+                print(f"Error detected in {trg_file} at line {i}.\n"
+                      f"==> Index {trg_label_index} points to {trg_label_value} instead of {src_label_index}:{src_label_value}")
+
+print("Detection terminated.")