First commit

Dockerfile

@@ -0,0 +1,21 @@
+FROM python:3.10-slim
+
+WORKDIR /app
+
+# System dependencies
+RUN apt-get update && apt-get install -y \
+    libgl1 \
+    tesseract-ocr \
+    && rm -rf /var/lib/apt/lists/*
+
+# Copy files
+COPY . .
+
+# Install Python dependencies
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Expose port (HF expects 7860)
+EXPOSE 7860
+
+# Run app
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

main.py

@@ -0,0 +1,41 @@
+from fastapi import FastAPI, UploadFile, File
+from fastapi.middleware.cors import CORSMiddleware
+import numpy as np
+import cv2
+import shutil
+import os
+import uuid
+from utils import process_image, process_video_stream
+
+app = FastAPI(title="DeepSight Lite API")
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.post("/predict-image")
+async def predict_image(file: UploadFile = File(...)):
+    contents = await file.read()
+    nparr = np.frombuffer(contents, np.uint8)
+    img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
+    results = process_image(img)
+    return {"status": "success", "data": results}
+
+@app.post("/predict-video")
+async def predict_video(file: UploadFile = File(...)):
+    temp_path = f"temp_{uuid.uuid4()}.mp4"
+    with open(temp_path, "wb") as buffer:
+        shutil.copyfileobj(file.file, buffer)
+    try:
+        detected_plates = process_video_stream(temp_path)
+        return {"status": "success", "plates": detected_plates}
+    finally:
+        if os.path.exists(temp_path):
+            os.remove(temp_path)
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=7860)

requirements.txt

@@ -0,0 +1,8 @@
+fastapi
+uvicorn
+opencv-python-headless
+numpy
+ultralytics
+easyocr
+torch
+torchvision

utils.py

@@ -0,0 +1,159 @@
+import cv2
+import numpy as np
+from ultralytics import YOLO
+import easyocr
+import os
+import uuid
+
+# =========================
+# Load YOLO model
+# =========================
+MODEL_PATH = "weights/best.pt"
+if not os.path.exists(MODEL_PATH):
+    print(f"❌ ERROR: Model weights not found at {MODEL_PATH}")
+
+model = YOLO(MODEL_PATH)
+
+# =========================
+# EasyOCR Init
+# =========================
+reader = easyocr.Reader(['en'], gpu=False)  # set gpu=False if needed
+
+# =========================
+# Debug folder
+# =========================
+DEBUG_DIR = "debug_crops"
+os.makedirs(DEBUG_DIR, exist_ok=True)
+
+
+# =========================
+# OCR FUNCTION (EasyOCR)
+# =========================
+def get_ocr_text(crop, crop_id):
+    try:
+        if crop is None or crop.size == 0:
+            return None, 0.0
+
+        crop = cv2.resize(crop, None, fx=2, fy=2, interpolation=cv2.INTER_CUBIC)
+        crop_rgb = cv2.cvtColor(crop, cv2.COLOR_BGR2RGB)
+
+        cv2.imwrite(f"{DEBUG_DIR}/{crop_id}.jpg", crop)
+
+        results = reader.readtext(crop_rgb)
+
+        if not results:
+            return None, 0.0
+
+        texts = []
+        confidences = []
+
+        for (bbox, text, conf) in results:
+            clean_text = "".join([c for c in text if c.isalnum()])
+            
+            if len(clean_text) >= 2:   # allow small parts like "L8"
+                texts.append(clean_text)
+                confidences.append(conf)
+
+        if not texts:
+            return None, 0.0
+
+        # 🔥 SORT by vertical position (top → bottom)
+        results_sorted = sorted(results, key=lambda x: min([p[1] for p in x[0]]))
+
+        final_text = ""
+        for (_, text, _) in results_sorted:
+            clean = "".join([c for c in text if c.isalnum()])
+            if len(clean) >= 2:
+                final_text += clean
+
+        avg_conf = sum(confidences) / len(confidences)
+
+        print(f"✅ OCR [{crop_id}] -> {final_text}")
+
+        return final_text, float(avg_conf)
+
+    except Exception as e:
+        print(f"OCR Error: {e}")
+        return None, 0.0
+
+# =========================
+# IMAGE PROCESSING
+# =========================
+def process_image(image):
+    print(f"\n[STEP 1] Running YOLO detection...")
+    results = model(image, imgsz=320, verbose=False)
+
+    plates = []
+
+    found_count = len(results[0].boxes)
+    print(f"[STEP 2] YOLO found {found_count} bounding boxes.")
+
+    for r in results:
+        if r.boxes:
+            for box in r.boxes.xyxy.cpu().numpy():
+                crop_id = f"plate_{uuid.uuid4().hex[:6]}"
+
+                x1, y1, x2, y2 = map(int, box)
+
+                # 🔥 IMPORTANT FIX: bigger padding
+                h, w, _ = image.shape
+                pad = 15
+
+                crop = image[
+                    max(0, y1 - pad):min(h, y2 + pad),
+                    max(0, x1 - pad):min(w, x2 + pad)
+                ]
+
+                print(f"[STEP 3] Processing {crop_id}...")
+                text, conf = get_ocr_text(crop, crop_id)
+
+                if text:
+                    plates.append({
+                        "text": text,
+                        "confidence": conf,
+                        "debug_id": crop_id
+                    })
+                else:
+                    print(f"   ⚠️ OCR failed for {crop_id}")
+
+    return plates
+
+
+# =========================
+# VIDEO PROCESSING
+# =========================
+def process_video_stream(video_path):
+    print(f"\n🎥 Processing video: {video_path}")
+
+    cap = cv2.VideoCapture(video_path)
+    tracked_plates = {}
+    final_results = []
+
+    while cap.isOpened():
+        ret, frame = cap.read()
+        if not ret:
+            break
+
+        results = model.track(frame, persist=True, imgsz=320, verbose=False)
+
+        if results[0].boxes.id is not None:
+            ids = results[0].boxes.id.int().cpu().tolist()
+            boxes = results[0].boxes.xyxy.cpu().numpy()
+
+            for box, tid in zip(boxes, ids):
+                if tid not in tracked_plates:
+                    x1, y1, x2, y2 = map(int, box)
+
+                    crop_id = f"track_{tid}"
+
+                    crop = frame[y1:y2, x1:x2]
+
+                    text, conf = get_ocr_text(crop, crop_id)
+
+                    if text and len(text) >= 5:
+                        print(f"✅ Detected Plate: {text}")
+                        tracked_plates[tid] = text
+                        final_results.append(text)
+
+    cap.release()
+    return list(set(final_results))

weights/best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf6cb0453d80644f55f3c75e8bd34787e53a4a445b3a5e976ce128d008241a8f
+size 5428314