Burn Image Models

This is a Rust crate for image models, inspired by the Python timm package.

Examples of loading pretrained model:

use burn::backend::Wgpu;
use bimm::cache::disk::DiskCacheConfig;
use bimm::models::resnet::{PREFAB_RESNET_MAP, ResNet};

let device = Default::default();

let prefab = PREFAB_RESNET_MAP.expect_lookup_prefab("resnet18");

let weights = prefab
    .expect_lookup_pretrained_weights("tv_in1k")
    .fetch_weights(&DiskCacheConfig::default())
    .expect("Failed to fetch weights");

let model: ResNet<Wgpu> = prefab
    .to_config()
    .to_structure()
    .init(&device)
    .load_pytorch_weights(weights)
    .expect("Failed to load weights")
    // re-head the model to 10 classes:
    .with_classes(10)
    // Enable (drop_block_prob) stochastic block drops for training:
    .with_stochastic_drop_block(0.2)
    // Enable (drop_path_prob) stochastic depth for training:
    .with_stochastic_path_depth(0.1);

• bimm::cache - weight loading cache.
• bimm::compat - future-porting burn mechanisms.
  • bimm::compat::activation_wrapper::Activation - activation layer abstraction wrapper.
  • bimm::compat::normalization_wrapper::Normalization - normalization layer abstraction wrapper.
• bimm::layers - reusable neural network modules.
  • bimm::layers::blocks - miscellaneous blocks.
    • bimm::layers::blocks::conv_norm::ConvNorm2d - Conv2d + BatchNorm2d block.
    • bimm::layers::blocks::cna::CNA2d - Conv2d + Normalization + Activation block.
  • bimm::layers::drop - dropout layers.
    • bimm::layers::drop::drop_block - 2d drop block / spatial dropout.
    • bimm::layers::drop::drop_path - drop path / stochastic depth.
  • bimm::layers::patching - patching layers.
    • bimm::layers::patching::patch_embed - 2d patch embedding layer.
• bimm::models - complete model families.
  • bimm::models::resnet - ResNet
  • bimm::models::swin - The SWIN Family.
    • bimm::models::swin::v2 - The SWIN-V2 Model.

Recent Changes

• 0.19.0
  • Bumped to track burn 0.19.0.
• 0.3.3
  • Preview of ResNet-18 support.
• 0.3.2
  • Fixed visibility for DropBlock3d / drop_block_3d support.
• 0.3.1
  • added DropBlock2d / drop_block_2d support.
• 0.2.0
  • bumped burn dependency to 0.18.0.