GeneMAN: Generalizable Single-Image 3D Human Reconstruction from Multi-Source Human Data
Paper • 2411.18624 • Published
_class_name string | _diffusers_version string | _name_or_path string | act_fn string | addition_embed_type null | addition_embed_type_num_heads int64 | addition_time_embed_dim null | attention_head_dim int64 | block_out_channels list | center_input_sample bool | class_embed_type null | class_embeddings_concat bool | conv_in_kernel int64 | conv_out_kernel int64 | cross_attention_dim int64 | cross_attention_norm null | decay float64 | down_block_types list | downsample_padding int64 | dual_cross_attention bool | encoder_hid_dim null | encoder_hid_dim_type null | flip_sin_to_cos bool | freq_shift int64 | in_channels int64 | inv_gamma float64 | layers_per_block int64 | mid_block_only_cross_attention null | mid_block_scale_factor int64 | mid_block_type string | min_decay float64 | norm_eps float64 | norm_num_groups int64 | num_attention_heads null | num_class_embeds null | only_cross_attention bool | optimization_step int64 | out_channels int64 | power float64 | projection_class_embeddings_input_dim null | resnet_out_scale_factor float64 | resnet_skip_time_act bool | resnet_time_scale_shift string | sample_size int64 | time_cond_proj_dim null | time_embedding_act_fn null | time_embedding_dim null | time_embedding_type string | timestep_post_act null | transformer_layers_per_block int64 | up_block_types list | upcast_attention bool | update_after_step int64 | use_ema_warmup bool | use_linear_projection bool |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
UNet2DConditionModel | 0.19.0.dev0 | silu | null | 64 | null | 8 | [
320,
640,
1280,
1280
] | false | null | false | 3 | 3 | 768 | null | 0.9999 | [
"CrossAttnDownBlock2D",
"CrossAttnDownBlock2D",
"CrossAttnDownBlock2D",
"DownBlock2D"
] | 1 | false | null | null | true | 0 | 4 | 1 | 2 | null | 1 | UNetMidBlock2DCrossAttn | 0 | 0.00001 | 32 | null | null | false | 500 | 4 | 0.666667 | null | 1 | false | default | 64 | null | null | null | positional | null | 1 | [
"UpBlock2D",
"CrossAttnUpBlock2D",
"CrossAttnUpBlock2D",
"CrossAttnUpBlock2D"
] | false | 0 | false | false |
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
GeneMAN: Generalizable Single-Image 3D Human Reconstruction from Multi-Source Human Data
Project Page: https://roooooz.github.io/GeneMAN/
Code Link: https://github.com/3DTopia/GeneMAN
Paper Link: http://arxiv.org/abs/2411.18624
Copy the pretrained_models folder into GeneMAN/pretrained_models.
Copy the tets folder into GeneMAN/extern.
- Downloads last month
- 55