File size: 1,222 Bytes
0ac03fd
aef0ec0
7aeddce
 
aef0ec0
0ac03fd
7aeddce
 
 
79c6af6
4310f73
b9a3ea6
 
 
7aeddce
 
4310f73
 
79c6af6
 
 
 
 
 
b9a3ea6
 
79c6af6
b9a3ea6
 
7aeddce
79c6af6
7aeddce
b9a3ea6
 
 
7aeddce
 
b9a3ea6
 
 
7aeddce
 
 
b9a3ea6
4310f73
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
---
license: apache-2.0
tags:
- unconditional-image-generation
datasets: skytnt/fbanimehq
---

# Model info

Project [fbanimegan](https://github.com/SkyTNT/fbanimegan)

### fbanime.pkl

StyleGan2 model trained with official [StyleGan3](https://github.com/NVlabs/stylegan3).
But I modified the code (networks_stylegan2.py and dataset.py) to support non-square resolutions.

FID: 1.4

### fbanime_fp32.pkl

fp32 version of fbanime.pkl

Note: Fp16 version (fbanime.pkl) only works on gpu. And fp32 version works on gpu and cpu.

### g_mapping.onnx

onnx format mapping network of fbanime_fp32.pkl

### g_synthesis.onnx

onnx format synthesis network of fbanime_fp32.pkl

### encoder.onnx

e4e model trained with [encoder4editing-stylegan3](https://github.com/yj7082126/encoder4editing-stylegan3).
I add support for official StyleGan2 model and change backbone to ResNet-34 in [restyle-encoder](https://github.com/yuval-alaluf/restyle-encoder).

### waifu_dect.onnx

YOLOv5 model trained with official [YOLOv5](https://github.com/ultralytics/yolov5)

# Usage

see [demo](https://huggingface.co/spaces/skytnt/full-body-anime-gan/blob/main/app.py)

# Dataset

[fbanimehq](https://huggingface.co/datasets/skytnt/fbanimehq) v2.0