Sentiodirector · abhiveer50 · Mar 18, 2026 · Mar 18, 2026
diff --git a/.DS_Store b/.DS_Store
diff --git a/README.md b/README.md
diff --git a/compressed_output.mp4 b/compressed_output.mp4
diff --git a/compression_report.html b/compression_report.html
@@ -0,0 +1 @@
+<html><body><h1>Sentio Mind Report</h1><p>Time: 71.68s</p><p>Reduction: 98.9%</p></body></html>
diff --git a/demo.mov b/demo.mov
diff --git a/segments_kept.json b/segments_kept.json
diff --git a/solution.py b/solution.py
@@ -0,0 +1,177 @@
+import cv2
+import json
+import base64
+import subprocess
+import time
+import numpy as np
+import os
+from pathlib import Path
+from PIL import Image
+from imagehash import phash
+
+# ---------------------------------------------------------------------------
+# CONFIG
+# ---------------------------------------------------------------------------
+VIDEO_IN               = Path("/Users/abhiveer/Downloads/Class_8_cctv_video_1.mov")
+VIDEO_OUT              = Path("compressed_output.mp4")
+REPORT_HTML_OUT        = Path("compression_report.html")
+SEGMENTS_JSON_OUT      = Path("segments_kept.json")
+
+PHASH_THRESHOLD        = 3      # Hamming distance threshold for imagehash.phash
+MOTION_KEEP_THRESH     = 0.15   # Standard threshold for motion
+CONTEXT_EVERY_SEC      = 3      # Step 4: Force frame every 3s
+OUTPUT_FPS             = 12     # Step 5: Output framerate
+OUTPUT_CRF             = 28     
+
+# ---------------------------------------------------------------------------
+# PERCEPTUAL HASH (Step 1)
+# ---------------------------------------------------------------------------
+
+def compute_phash(frame: np.ndarray):
+    """Uses imagehash library as per your logic."""
+    # Small grayscale for speed
+    gray_small = cv2.resize(cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY), (160, 90))
+    return phash(Image.fromarray(gray_small))
+
+def phash_similarity(h1, h2) -> float:
+    """Returns Hamming distance for direct comparison."""
+    if h1 is None or h2 is None: return 999
+    return h1 - h2
+
+# ---------------------------------------------------------------------------
+# FACE PRESENCE CHECK (Step 3)
+# ---------------------------------------------------------------------------
+
+def has_face(frame: np.ndarray, cascade) -> bool:
+    """Detection on a small proxy for speed."""
+    gray_small = cv2.resize(cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY), (160, 90))
+    faces = cascade.detectMultiScale(gray_small, 1.2, 3)
+    return len(faces) > 0
+
+# ---------------------------------------------------------------------------
+# THUMBNAIL HELPER (For JSON Contract)
+# ---------------------------------------------------------------------------
+
+def frame_to_b64_thumb(frame: np.ndarray) -> str:
+    h, w = frame.shape[:2]
+    thumb = cv2.resize(frame, (200, int(h * 200 / w)))
+    _, buf = cv2.imencode(".jpg", thumb, [cv2.IMWRITE_JPEG_QUALITY, 60])
+    return base64.b64encode(buf).decode("utf-8")
+
+# ---------------------------------------------------------------------------
+# VIDEO WRITING (Step 5 - Mac Accelerated)
+# ---------------------------------------------------------------------------
+
+def write_frames_to_video(kept_frames: list, output_path: Path, fps: float, frame_size: tuple):
+    temp_raw = "temp_raw.mp4"
+    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
+    out = cv2.VideoWriter(temp_raw, fourcc, 12, frame_size)
+
+    for f in kept_frames:
+        out.write(f)
+    out.release()
+
+    # Hardware Acceleration for Mac (The <60s secret)
+    subprocess.run([
+        "ffmpeg", "-y", "-i", temp_raw, 
+        "-vcodec", "h264_videotoolbox", 
+        "-b:v", "2M", str(output_path)
+    ])
+    if os.path.exists(temp_raw): os.remove(temp_raw)
+
+# ---------------------------------------------------------------------------
+# MAIN PIPELINE
+# ---------------------------------------------------------------------------
+
+if __name__ == "__main__":
+    t_start = time.time()
+    cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
+
+    cap = cv2.VideoCapture(str(VIDEO_IN))
+    if not cap.isOpened():
+        print("❌ Could not open video."); exit()
+
+    total_in = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+    fps_in = cap.get(cv2.CAP_PROP_FPS) or 25.0
+    fw, fh = int(cap.get(3)), int(cap.get(4))
+    orig_mb = VIDEO_IN.stat().st_size / 1_000_000
+
+    kept_frames, segments = [], []
+    last_kept_hash, last_kept_t = None, -999.0
+    cur_seg = None
+    disc_dup, disc_stat = 0, 0
+
+    print(f"🚀 Processing {total_in} frames (720p Optimized)...")
+
+    frame_idx = 0
+    while True:
+        ret, frame = cap.read()
+        if not ret: break
+
+        # Speed Optimization: Process 1 out of every 2 frames
+        if frame_idx % 2 != 0:
+            frame_idx += 1
+            continue
+
+        ts = frame_idx / fps_in
+        curr_hash = compute_phash(frame)
+
+        keep, reason = False, ""
+
+        # 1. pHash Check
+        if last_kept_hash is None or phash_similarity(curr_hash, last_kept_hash) > PHASH_THRESHOLD:
+            keep, reason = True, "perceptual_hash"
+
+        # 3. Face Check (If pHash didn't trigger)
+        if not keep and has_face(frame, cascade):
+            keep, reason = True, "face_detected"
+
+        # 4. Context Check (Heartbeat every 3s)
+        if not keep and (ts - last_kept_t) >= CONTEXT_EVERY_SEC:
+            keep, reason = True, "scene_continuity"
+
+        if keep:
+            kept_frames.append(frame.copy())
+            last_kept_hash = curr_hash
+            last_kept_t = ts
+
+            # Segment Tracking for JSON contract
+            if cur_seg is None or (ts - cur_seg["end_sec"]) > 2.0 or reason != cur_seg["reason_kept"]:
+                if cur_seg: segments.append(cur_seg)
+                cur_seg = {
+                    "segment_id": len(segments) + 1, "start_sec": round(ts, 2), "end_sec": round(ts, 2),
+                    "frames_in_segment": 1, "reason_kept": reason,
+                    "thumbnail_b64": frame_to_b64_thumb(frame)
+                }
+            else:
+                cur_seg["end_sec"] = round(ts, 2)
+                cur_seg["frames_in_segment"] += 1
+
+        frame_idx += 1
+
+    if cur_seg: segments.append(cur_seg)
+    cap.release()
+
+    print("📦 Encoding video...")
+    write_frames_to_video(kept_frames, VIDEO_OUT, OUTPUT_FPS, (fw, fh))
+
+    # Deliverables
+    proc_time = time.time() - t_start
+    comp_mb = VIDEO_OUT.stat().st_size / 1_000_000
+
+    stats = {
+        "original_size_mb": round(orig_mb, 2),
+        "compressed_size_mb": round(comp_mb, 2),
+        "reduction_pct": round((1 - comp_mb/orig_mb)*100, 1),
+        "processing_time_sec": round(proc_time, 2),
+        "segments": segments
+    }
+
+    with open(SEGMENTS_JSON_OUT, 'w') as f:
+        json.dump(stats, f, indent=4)
+
+    # HTML Report
+    html = f"<html><body><h1>Sentio Mind Report</h1><p>Time: {proc_time:.2f}s</p><p>Reduction: {stats['reduction_pct']}%</p></body></html>"
+    with open(REPORT_HTML_OUT, "w") as f: f.write(html)
+
+    print(f"✅ Done in {proc_time:.2f}s | Reduction: {stats['reduction_pct']}%")
diff --git a/video_compression.json b/video_compression.json
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		<html><body><h1>Sentio Mind Report</h1><p>Time: 71.68s</p><p>Reduction: 98.9%</p></body></html>