Some README.md changes + prepare_dataset.py

README.md

@@ -10,6 +10,12 @@
 pip install ultralytics albumentations opencv-python tqdm pyyaml torch
 ```
 
+or using conda
+
+```bash
+conda env create -f enviroment.yml
+``
+
 ---
 
 ## Step-by-step
@@ -31,7 +37,7 @@ Put your raw product photos into `dataset/images/train/`.
 Use **Roboflow** (recommended) or **LabelImg** to draw bounding boxes and export in YOLO format:
 
 - [Roboflow](https://roboflow.com) — free, web-based, exports directly to YOLO format
-- LabelImg: `pip install labelImg && labelImg`
+- LabelImg: `pip install labelImg && labelImg` or download it from [here](https://github.com/tzutalin/labelImg/files/2638199/windows_v1.8.1.zip)
 
 Each image needs a matching `.txt` in `dataset/labels/train/`:
 
@@ -127,4 +133,4 @@ With 4 original images + augmentation and 13 classes:
 - **mAP50 ≈ 0.65–0.80** (good for production)
 - **Inference speed** ≈ 15–30ms/image on CPU, <5ms on GPU
 
-Adding even 2–3 more labeled images per class typically pushes mAP above 0.85.
+Adding even 2–3 more labeled images per class typically pushes mAP above 0.85.

enviroment.yml

@@ -0,0 +1,18 @@
+name: yolo-coffee-object-detection
+channels:
+  - conda-forge
+dependencies:
+  - python=3.10
+  - ultralytics
+  - albumentations
+  - opencv
+  - tqdm
+  - pyyaml
+  #- torch
+  #- numpy
+  #- opencv-contrib-python
+  #- scikit-learn # @Todo: Remove it later if we don't need it.
+  - pip
+  - pip:
+      - fastapi
+      - uvicorn[standard]

prepare_dataset.py

@@ -0,0 +1,107 @@
+"""
+STEP 1 ─ prepare_dataset.py
+===========================
+Run this FIRST. It creates the required folder structure and
+a dataset.yaml that YOLOv8 expects.
+
+Usage:
+    python prepare_dataset.py
+
+After running. your folder tree will look like this:
+    dataset/
+      images/
+        train/   <- put your labeled images here.
+        val/     <- validation split (auto-filled by augment.py)
+
+      labels/
+        train/   <- put your .txt label files here.
+        val/     <- auto-generated.
+    dataset.yaml <- auto-generated.
+
+Label format (one .txt per image, same name):
+    <class_id> <cx> <cy> <w> <h> (all values 0.0-1.0, normalised)
+
+Recommended free labeling tools:
+    • Roboflow   https://roboflow.com        (easiest, web-based)
+    • LabelImg   pip install labelImg        (desktop, saves YOLO format directly)
+"""
+import os
+import yaml
+
+# ── Edit this list to match your product classes ────────
+CLASS_NAMES = [
+        "1001",
+        "Aroma",
+        "Aroma Gold",
+        "Bonal",
+        "Facto",
+        "Famico",
+        "Famico Exclusive",
+        "Fegalo",
+        "Gosto",
+        "Molino",
+        "Oscar",
+        "Primo",
+        "Siglo"
+]
+
+DATASET_ROOT = "dataset"
+
+DIRS = [
+    f"{DATASET_ROOT}/images/train",
+    f"{DATASET_ROOT}/images/val",
+    f"{DATASET_ROOT}/labels/train",
+    f"{DATASET_ROOT}/labels/val",
+]
+
+def Structure_Create():
+    for d in DIRS:
+        os.makedirs(d, exist_ok=True)
+    print("[1/3] Folder structure created:")
+
+    for d in DIRS:
+        print("         {d}/")
+
+def YAML_Write():
+    cfg = {
+            "path": os.path.abspath(DATASET_ROOT),
+            "train": "images/train",
+            "val": "images/val",
+            "nc": len(CLASS_NAMES),
+            "names": CLASS_NAMES,
+    }
+
+    yaml_path = "dataset.yml"
+    with open(yaml_path, "w") as f:
+        yaml.dump(cfg, f, default_flow_style=False, sort_keys=False)
+
+    print("[2/3] dataset.yaml written -> {yaml_path}")
+    return yaml_path
+
+def Print_NextSetups():
+    print("""
+[3/3] NEXT STEPS
+─────────────────────────────────────────────────────────────
+1. Put your RAW images (one per product view) into:
+       dataset/images/train/
+
+2. Label them using Roboflow or LabelImg (YOLO format).
+   Each image  →  a matching .txt in dataset/labels/train/
+   Example label file  (famico_veloute_01.txt):
+       0 0.512 0.483 0.320 0.541
+
+3. Run the augmentation script:
+       python 2_augment.py
+
+4. Train:
+       python 3_train.py
+
+5. Detect on new scenes:
+       python 4_detect.py --source scene.jpg
+─────────────────────────────────────────────────────────────
+    """)
+
+if __name__ == "__main__":
+    Structure_Create()
+    YAML_Write()
+    Print_NextSetups()