dearsky/Magic123
1
0
Fork 0
Code Issues 6 Pull Requests Actions Packages Projects Releases Wiki Activity

added eigen requriement and installation

Browse Source
This commit is contained in:
Guocheng Qian
2023-08-08 18:50:42 +00:00
parent 4ad512ad41
commit a99ac494bc
7 changed files with 12 additions and 178 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
readme.md
View File

@@ -35,12 +35,19 @@ Official PyTorch Implementation of Magic123: One Image to High-Quality 3D Object
# Install
We only test on Ubuntu system. Make sure git, wget, Eigen are installed.
```
apt update && apt upgrade
apt install git wget libeigen3-dev -y
```
### Install Environment
```bash
source install.sh
```
Note: in this `install.sh`, we use python venv by default. If you prefer conda, uncomment the conda and comment venv in the file and run the same command.
### Download pre-trained models
@@ -63,7 +70,7 @@ source install.sh
# Usage
## Preprocess [Optional]
We have included all preprocessed files in `./data` directory. Preprocessing is only necessary if you want to test on your own examples.
We have included all preprocessed files in `./data` directory. Preprocessing is only necessary if you want to test on your own examples. Takes seconds.
### Step1: Extract depth
```
@@ -72,7 +79,7 @@ python preprocess_image.py --path /path/to/image
### Step 2: Textural inversion [Optional]
Magic123 uses the defualt [textural inversion](https://huggingface.co/docs/diffusers/training/text_inversion) from diffuers, which consumes around 2.5 hours on a 32G V100. If you do not want to spend time in this textural inversion, you can: (1) study whether there is other faster textural inversion; or (2) do not use textural inversion in the loss of texture and shape consistencies. To run textural inversion:
Magic123 uses the defualt [textural inversion](https://huggingface.co/docs/diffusers/training/text_inversion) from diffuers, which consumes around 2 hours on a 32G V100. If you do not want to spend time in this textural inversion, you can: (1) study whether there is other faster textural inversion; or (2) do not use textural inversion in the loss of texture and shape consistencies. To run textural inversion:
```
bash scripts/texural_inversion/textural_inversion.sh $GPU_IDX runwayml/stable-diffusion-v1-5 /path/to/example/rgba.png /path/to/save $token_name $init_token --max_train_steps 5000