Knowing commited on
Commit
29ee0cb
·
verified ·
1 Parent(s): 58c84ec

Upload RE10K DL3DV 2-view Hub-format checkpoint

Browse files
re10k-dl3dv-2view/README.md ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: pytorch
3
+ pipeline_tag: image-to-3d
4
+ tags:
5
+ - 3d-gaussian-splatting
6
+ - novel-view-synthesis
7
+ - feed-forward
8
+ - vggt
9
+ - f4splat
10
+ ---
11
+
12
+ # F4Splat
13
+
14
+ This is the F4Splat RE10K+DL3DV 2-view checkpoint converted to Hugging Face
15
+ Hub format. It is intended to be loaded with `src.model.f4splat_hub.F4SplatHubModel`.
16
+
17
+ ```python
18
+ import torch
19
+ from src.model.f4splat_hub import F4SplatHubModel
20
+
21
+ model = F4SplatHubModel.from_pretrained("Knowing/F4Splat")
22
+ model = model.eval().cuda()
23
+ ```
24
+
25
+ The model expects 2-16 RGB context images resized to 256x256 and normalized to
26
+ `[0, 1]`. The local Gradio demo in this repository exposes budget control and
27
+ interactive PLY export.
28
+
29
+ Project page: https://mlvlab.github.io/F4Splat/
30
+ Paper: https://arxiv.org/abs/2603.21304
re10k-dl3dv-2view/config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "decoder": {
3
+ "background_color": [
4
+ 0.0,
5
+ 0.0,
6
+ 0.0
7
+ ],
8
+ "name": "splatting_cuda"
9
+ },
10
+ "encoder": {
11
+ "backbone": {
12
+ "align_corners": false,
13
+ "gs_param_dim": 256,
14
+ "gs_sh_degree": 3,
15
+ "head_mode": "pcd",
16
+ "multi_scale_list": [
17
+ 0.0
18
+ ],
19
+ "name": "vggt",
20
+ "pretrained_model": null
21
+ },
22
+ "density_control": {
23
+ "gs_param_dim": 256,
24
+ "mean_dim": 32,
25
+ "name": "density_control_module",
26
+ "num_level": -1
27
+ },
28
+ "multi_scale_list": [
29
+ 0.25,
30
+ 0.5,
31
+ 1.0
32
+ ],
33
+ "name": "f4splat"
34
+ },
35
+ "far": 100.0,
36
+ "image_size": [
37
+ 256,
38
+ 256
39
+ ],
40
+ "model_type": "f4splat",
41
+ "near": 0.1,
42
+ "variant": "re10k_dl3dv_2view"
43
+ }
re10k-dl3dv-2view/conversion_report.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "checkpoint": "downloaded_checkpoints/hf/Knowing_F4Splat_private_weight/outputs/full/re10k_dl3dv_2view/DL3DV_RE10k_FULL_2view/checkpoints/epoch_0-step_18750.ckpt",
3
+ "output_dir": "hub/f4splat/re10k-dl3dv-2view",
4
+ "repo_id": "Knowing/F4Splat",
5
+ "num_tensors": 1429,
6
+ "missing": [],
7
+ "unexpected": []
8
+ }
re10k-dl3dv-2view/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:646aea87b9b1b72c90c2c000ae73b02c1c16c4ea6d24dd5defa403a960ff9c89
3
+ size 4771611976