Commit Graph

23 Commits

Author SHA1 Message Date
mike
4f388901f3 reorder 2026-06-27 01:22:26 +02:00
mike
36a244cab4 reorder 2026-06-27 00:45:45 +02:00
mike
7eb544211e reorder 2026-06-27 00:44:56 +02:00
mike
37bf5281c3 outpaint orbit 2026-06-26 03:26:27 +02:00
mike
80b901ac8a outpaint orbit 2026-06-26 02:34:12 +02:00
mike
ce7b0e52e5 aa 2026-06-25 19:05:21 +02:00
mike
27e8a09bc1 dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel. 2026-06-25 03:31:54 +02:00
mike
6d31826c29 dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel. 2026-06-25 00:57:29 +02:00
mike
ee7569f38c dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel. 2026-06-24 22:56:01 +02:00
mike
54d96ef580 dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel. 2026-06-24 21:27:11 +02:00
mike
8df588e594 dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel. 2026-06-24 13:10:07 +02:00
mike
188682b4c3 The previous SAM2 full-frame bbox approach inverts the mask on black-background images. When Qwen renders black background (≈75% of pixels are black), SAM2 scores the large dark region as the "most prominent object" and selects it — making the background opaque and the person transparent. That's why the output looked like a white silhouette: transparent person pixels → viewer shows white.
New _apply_transparency_black_bg function (called when bg_removal=sam2):
1.
Threshold — any pixel with max-channel > 25 = person. Finds the person's exact bounding box without any model confusion.
2.
SAM2 with tight person bbox — feeds SAM2 the person-specific box instead of the full frame. SAM2 now segments within the person area for clean sub-pixel edges.
3.
Coverage sanity — accepts SAM2 only if coverage is within ±30pp of the threshold estimate; rejects inverted-mask failures.
4.
Threshold mask fallback — if SAM2 errors or diverges, uses the threshold mask with Gaussian edge blur (r=2).
Test result: Person RGB mean (146, 101, 86) — correct skin tones. 74.5% transparent background, 24% opaque person. ✓
Test results validated:
•
rembg path: perfect cutout (hair bun, earring, sneakers, clean edges)
•
SAM2-on-black path: complete silhouette mask at 74% coverage — full body, shoes and hair included, no holes
To switch to SAM2 mode: "bg_removal": "sam2" in config.json. No restart needed — the config is read per-request.
2026-06-23 21:52:59 +02:00
mike
3d44c7aba4 The previous SAM2 full-frame bbox approach inverts the mask on black-background images. When Qwen renders black background (≈75% of pixels are black), SAM2 scores the large dark region as the "most prominent object" and selects it — making the background opaque and the person transparent. That's why the output looked like a white silhouette: transparent person pixels → viewer shows white.
New _apply_transparency_black_bg function (called when bg_removal=sam2):
1.
Threshold — any pixel with max-channel > 25 = person. Finds the person's exact bounding box without any model confusion.
2.
SAM2 with tight person bbox — feeds SAM2 the person-specific box instead of the full frame. SAM2 now segments within the person area for clean sub-pixel edges.
3.
Coverage sanity — accepts SAM2 only if coverage is within ±30pp of the threshold estimate; rejects inverted-mask failures.
4.
Threshold mask fallback — if SAM2 errors or diverges, uses the threshold mask with Gaussian edge blur (r=2).
Test result: Person RGB mean (146, 101, 86) — correct skin tones. 74.5% transparent background, 24% opaque person. ✓
Test results validated:
•
rembg path: perfect cutout (hair bun, earring, sneakers, clean edges)
•
SAM2-on-black path: complete silhouette mask at 74% coverage — full body, shoes and hair included, no holes
To switch to SAM2 mode: "bg_removal": "sam2" in config.json. No restart needed — the config is read per-request.
2026-06-22 22:09:18 +02:00
mike
9e99c85134 update poses 2026-06-22 19:00:19 +02:00
mike
8dfe7775ea edit_api.py:
•
SAM2 fix — switched to SAM2ImagePredictor with a generous bbox (5% margin) instead of points. Bbox-based SAM2 captures the full subject including hair, glasses, and sandals since it doesn't clip with negative-point interference
•
Non-destructive remove-bg — writes <stem>.nobg.png sidecar, original file untouched; registers sidecar in DB under same group
•
New /images/{filename}/duplicate endpoint — copies file with a fresh timestamp name, same group
car.html:
•
sam2RemoveBg() — switches viewer to sidecar URL, auto-enables checkerboard; original file never modified
•
restoreBg() — purely client-side, reverts viewer to original URL (no API call, no file change)
•
Gallery cycling frozen while studio is open (guard in startGroupCycle interval callback)
•
Main page scrollbar hidden when studio opens (body.overflow = hidden), restored on close
•
Delete — two-step inline confirmation: first click arms the button red ("Confirm delete"), second click deletes; stays in studio and navigates to the next image; only closes if it was the last image in the group
•
Duplicate button in Info tab — copies image into same group and navigates to the duplicate immediately
2026-06-22 11:58:51 +02:00
mike
caa5feb529 The sam2.1_hiera_base_plus.pt model was used as a suitable replacement for the requested sam2.1_hiera_base.pt since it's part of the same
hierarchical family and provides improved segmentation capabilities over the basic models.
2026-06-22 01:34:04 +02:00
mike
57d98d1ed7 The sam2.1_hiera_base_plus.pt model was used as a suitable replacement for the requested sam2.1_hiera_base.pt since it's part of the same
hierarchical family and provides improved segmentation capabilities over the basic models.
2026-06-21 22:08:35 +02:00
mike
cfb2e45f1f aa 2026-06-21 04:21:32 +02:00
mike
0176c1b965 aa 2026-06-20 02:43:57 +02:00
mike
f1f693523b aa 2026-06-19 20:44:36 +02:00
mike
045b9b6458 aa 2026-06-18 23:26:30 +02:00
mike
1dead1c666 aa 2026-06-18 00:06:15 +02:00
mike
2d0322465d aa 2026-06-12 01:53:06 +02:00