39ba5ddf76
Schema changes: dev.chunks->dev.chunk, remove old_chunk_id/chunk_index Correction: asr-1.json format, generate/apply scripts API: 37/37 endpoints fixed and tested Docs: HANDOVER_V2.0.md for M4
104 lines
3.9 KiB
Python
104 lines
3.9 KiB
Python
#!/opt/homebrew/bin/python3.11
|
|
"""
|
|
Test Grounding DINO Large on stamps, envelopes, passports, letters.
|
|
"""
|
|
import json, os, time, cv2, torch
|
|
from PIL import Image
|
|
from transformers import AutoProcessor, AutoModelForZeroShotObjectDetection
|
|
|
|
MODEL_PATH = "/Users/accusys/momentry_core_0.1/models/gun/grounding-dino-large-hf"
|
|
VIDEO = "/Users/accusys/momentry/var/sftpgo/data/demo/Charade (1963) Cary Grant & Audrey Hepburn \uff5c Comedy Mystery Romance Thriller \uff5c Full Movie.mp4"
|
|
OUTPUT_DIR = "/Users/accusys/momentry/output_dev/zero_shot_objects"
|
|
os.makedirs(OUTPUT_DIR, exist_ok=True)
|
|
|
|
# Timepoints per object type
|
|
TESTS = [
|
|
# (label, time_sec, prompts)
|
|
("stamp_001", 429, ["stamp", "postage stamp"]),
|
|
("stamp_002", 691, ["stamp", "envelope", "letter"]),
|
|
("stamp_003", 5443, ["stamp", "envelope"]),
|
|
("stamp_004", 5500, ["stamp"]),
|
|
("stamp_005", 5506, ["stamp"]),
|
|
("envelope_001", 5443, ["envelope"]),
|
|
("envelope_002", 5467, ["envelope"]),
|
|
("envelope_003", 5786, ["envelope"]),
|
|
("passport_001", 762, ["passport", "identification"]),
|
|
("passport_002", 3491, ["passport", "identification"]),
|
|
("passport_003", 5054, ["passport"]),
|
|
("letter_001", 691, ["letter", "envelope"]),
|
|
("letter_002", 5434, ["letter", "envelope"]),
|
|
("letter_003", 5783, ["letter", "stamp"]),
|
|
]
|
|
|
|
print(f"Loading Large model...")
|
|
t0 = time.time()
|
|
processor = AutoProcessor.from_pretrained(MODEL_PATH)
|
|
model = AutoModelForZeroShotObjectDetection.from_pretrained(MODEL_PATH)
|
|
device = "mps" if torch.backends.mps.is_available() else "cpu"
|
|
model.to(device)
|
|
print(f"Loaded in {time.time()-t0:.1f}s, device={device}")
|
|
|
|
cap = cv2.VideoCapture(VIDEO)
|
|
fps = cap.get(cv2.CAP_PROP_FPS) or 25.0
|
|
|
|
results = {}
|
|
t_infer = time.time()
|
|
|
|
for label, t_sec, prompts in TESTS:
|
|
cap.set(cv2.CAP_PROP_POS_FRAMES, int(t_sec * fps))
|
|
ret, frame = cap.read()
|
|
if frame is None: continue
|
|
|
|
img = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
|
|
key = f"{label}_{t_sec}s"
|
|
results[key] = {"time": t_sec, "time_str": f"{t_sec//60}:{t_sec%60:02d}", "prompts": {}}
|
|
|
|
for prompt in prompts:
|
|
inputs = processor(images=img, text=f"{prompt}.", return_tensors="pt").to(device)
|
|
with torch.no_grad():
|
|
outputs = model(**inputs)
|
|
target = torch.tensor([img.size[::-1]])
|
|
dets = processor.post_process_grounded_object_detection(
|
|
outputs, threshold=0.1, target_sizes=target
|
|
)[0]
|
|
|
|
det_list = []
|
|
for i in range(len(dets["boxes"])):
|
|
det_list.append({
|
|
"bbox": [round(v, 1) for v in dets["boxes"][i].tolist()],
|
|
"score": round(dets["scores"][i].item(), 3),
|
|
})
|
|
results[key]["prompts"][prompt] = det_list
|
|
|
|
# Save annotated frame
|
|
if det_list:
|
|
cv2_img = frame.copy()
|
|
for d in det_list:
|
|
x1, y1, x2, y2 = [int(v) for v in d["bbox"]]
|
|
cv2.rectangle(cv2_img, (x1, y1), (x2, y2), (0, 255, 0), 2)
|
|
cv2.putText(cv2_img, f"{prompt} {d['score']:.2f}", (x1, y1-5),
|
|
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
|
|
cv2.imwrite(os.path.join(OUTPUT_DIR, f"{label}_{t_sec}s_{prompt}.jpg"), cv2_img,
|
|
[cv2.IMWRITE_JPEG_QUALITY, 85])
|
|
|
|
cap.release()
|
|
elapsed = time.time() - t_infer
|
|
|
|
# Summary
|
|
print(f"\n{'='*60}")
|
|
print(f"Results ({elapsed:.0f}s)")
|
|
print(f"{'='*60}")
|
|
for key, data in sorted(results.items()):
|
|
found = [p for p, dets in data["prompts"].items() if dets]
|
|
if found:
|
|
best = max(
|
|
((p, d["score"]) for p, dets in data["prompts"].items() for d in dets),
|
|
key=lambda x: x[1]
|
|
)
|
|
print(f" {data['time_str']} {key:20s} ✅ {best[1]:.3f} ({best[0]})")
|
|
else:
|
|
print(f" {data['time_str']} {key:20s} ❌ none")
|
|
|
|
json.dump(results, open(os.path.join(OUTPUT_DIR, "results.json"), "w"), indent=2)
|
|
print(f"\nScreenshots saved to {OUTPUT_DIR}/")
|