We present FlatFormer, an efficient ViT architecture for large-scale point cloud analysis.

EIE proposed to accelerate pruned and compressed neural networks, exploiting weight sparsity, activation sparsity, and 4-bit weight-sharing in neural network accelerators.

BEVFusion unifies multi-modal features in the shared bird’s-eye view (BEV) representation space, which nicely preserves both geometric and semantic information. It establishes the new state of the art on nuScenes, achieving 1.3% higher mAP and NDS on 3D object detection and 13.6% higher mIoU on BEV map segmentation with 1.9x lower computation cost.

An engine that selectively performs computations at the edited regions to accelerate image editing applications.

TorchSparse is a high-performance computing library for efficient 3D sparse convolution. This library aims at accelerating sparse computation in 3D, in particular the Sparse Convolution operation.

In MCUNetV3, we enable on-device training under 256KB memory, using less than 1/1000 memory of PyTorch while matching the accuracy on the visual wake words application using system-algorithm co-design.