DragMesh-3-Scene / scenesmith_full /scripts /test_material_retrieval.py
Neptune-T's picture
Add files using upload-large-folder tool
733b9f3 verified
Raw History Blame Contribute Delete
5.53 kB
#!/usr/bin/env python3
"""Test AmbientCG material retrieval with CLIP embeddings.
This script tests CLIP-based retrieval by querying with text and displaying
retrieved material information.
Usage:
python scripts/test_material_retrieval.py \
--materials-dir data/materials \
--query "red brick wall" \
--top-k 5
# Save preview images to output directory
python scripts/test_material_retrieval.py \
--materials-dir data/materials \
--query "wooden floor" \
--top-k 3 \
--output-path material_test
"""
import argparse
import logging
import shutil
from pathlib import Path
import numpy as np
import yaml
from scenesmith.agent_utils.clip_embeddings import (
compute_clip_similarities,
get_text_embedding,
)
console_logger = logging.getLogger(__name__)
def load_material_data(
embeddings_path: Path,
) -> tuple[np.ndarray, list[str], dict]:
"""Load material embeddings and metadata.
Args:
embeddings_path: Path to embeddings directory.
Returns:
Tuple of (embeddings, material_ids, metadata_dict).
"""
embeddings = np.load(embeddings_path / "clip_embeddings.npy")
with open(embeddings_path / "embedding_index.yaml") as f:
material_ids = yaml.safe_load(f)
with open(embeddings_path / "metadata_index.yaml") as f:
metadata = yaml.safe_load(f)
return embeddings, material_ids, metadata
def retrieve_top_k(
query: str,
embeddings: np.ndarray,
material_ids: list[str],
top_k: int,
) -> list[tuple[str, float]]:
"""Retrieve top-k materials by CLIP similarity.
Args:
query: Text query.
embeddings: Embedding matrix (N x D).
material_ids: List of material IDs corresponding to embeddings.
top_k: Number of results to return.
Returns:
List of (material_id, similarity_score) tuples.
"""
text_embedding = get_text_embedding(query)
all_indices = list(range(len(material_ids)))
similarities = compute_clip_similarities(
query_embedding=text_embedding,
embeddings=embeddings,
indices=all_indices,
)
sorted_results = sorted(similarities.items(), key=lambda x: x[1], reverse=True)
results = []
for idx, score in sorted_results[:top_k]:
results.append((material_ids[idx], score))
return results
def main() -> None:
"""Main entry point."""
parser = argparse.ArgumentParser(
description="Test AmbientCG material retrieval with CLIP embeddings."
)
parser.add_argument(
"--materials-dir",
type=Path,
required=True,
help="Path to materials directory (with embeddings/ subdirectory).",
)
parser.add_argument(
"--query",
type=str,
required=True,
help="Text query for retrieval.",
)
parser.add_argument(
"--top-k",
type=int,
default=5,
help="Number of results to retrieve (default: 5).",
)
parser.add_argument(
"--output-path",
type=Path,
default=None,
help="Optional path to save preview images.",
)
args = parser.parse_args()
# Set up logging.
logging.basicConfig(level=logging.INFO, format="%(levelname)s: %(message)s")
# Validate paths.
embeddings_path = args.materials_dir / "embeddings"
if not embeddings_path.exists():
console_logger.error(f"Embeddings not found: {embeddings_path}")
console_logger.error("Run compute_ambientcg_embeddings.py first.")
return
# Load data.
console_logger.info("Loading material embeddings...")
embeddings, material_ids, metadata = load_material_data(embeddings_path)
console_logger.info(
f"Loaded {len(material_ids)} materials with {embeddings.shape[1]}D embeddings"
)
# Retrieve top-k.
console_logger.info(f"\nQuerying: '{args.query}'")
results = retrieve_top_k(
query=args.query,
embeddings=embeddings,
material_ids=material_ids,
top_k=args.top_k,
)
# Display results.
console_logger.info(f"\nTop {len(results)} results:")
console_logger.info("=" * 60)
for i, (material_id, score) in enumerate(results, 1):
meta = metadata.get(material_id, {})
category = meta.get("category", "Unknown")
tags = meta.get("tags", [])
tags_str = ", ".join(tags[:8]) # Show first 8 tags.
if len(tags) > 8:
tags_str += f", ... (+{len(tags) - 8} more)"
console_logger.info(f"\n{i}. {material_id} (score: {score:.4f})")
console_logger.info(f" Category: {category}")
console_logger.info(f" Tags: {tags_str}")
# Copy preview images if output path specified.
if args.output_path:
args.output_path.mkdir(parents=True, exist_ok=True)
console_logger.info(f"\nCopying preview images to: {args.output_path}")
for i, (material_id, score) in enumerate(results, 1):
preview_src = args.materials_dir / material_id / f"{material_id}.png"
if preview_src.exists():
preview_dst = args.output_path / f"rank{i:02d}_{material_id}.png"
shutil.copy(preview_src, preview_dst)
console_logger.info(f" Copied: {preview_dst.name}")
else:
console_logger.warning(f" Preview not found: {preview_src}")
console_logger.info(f"\nDone! Results saved to: {args.output_path}")
if __name__ == "__main__":
main()