Buckets:
| """Extract matched MJHQ real images for the sampled prompts (FID-vs-real), resized to RES. | |
| Usage: python3 scripts/33_extract_ref.py [RES] [N] | |
| """ | |
| import sys, json, os, io, zipfile | |
| from PIL import Image | |
| RES = int(sys.argv[1]) if len(sys.argv) > 1 else 512 | |
| N = int(sys.argv[2]) if len(sys.argv) > 2 else 10**9 | |
| sel = json.load(open('outputs/eval/prompts.json'))[:N] | |
| z = zipfile.ZipFile('data/mjhq_raw/mjhq30k_imgs.zip') | |
| out = 'outputs/eval/imgs/mjhq_ref'; os.makedirs(out, exist_ok=True) | |
| n = 0 | |
| for d in sel: | |
| arc = f"{d['category']}/{d['id']}.jpg" | |
| try: | |
| b = z.read(arc) | |
| except KeyError: | |
| print("missing", arc); continue | |
| Image.open(io.BytesIO(b)).convert('RGB').resize((RES, RES), Image.LANCZOS).save( | |
| f"{out}/{d['idx']:05d}.jpg", quality=95) | |
| n += 1 | |
| print(f"extracted {n} ref imgs @ {RES} -> {out}") | |
Xet Storage Details
- Size:
- 838 Bytes
- Xet hash:
- 2ccf895c87d1b4178c06ab7eff6347292de5e14af3c29a5d9c98ea9ee188ba15
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.