Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

PTQ4SAM: Post-Training Quantization for Segment Anything

About

Segment Anything Model (SAM) has achieved impressive performance in many computer vision tasks. However, as a large-scale model, the immense memory and computation costs hinder its practical deployment. In this paper, we propose a post-training quantization (PTQ) framework for Segment Anything Model, namely PTQ4SAM. First, we investigate the inherent bottleneck of SAM quantization attributed to the bimodal distribution in post-Key-Linear activations. We analyze its characteristics from both per-tensor and per-channel perspectives, and propose a Bimodal Integration strategy, which utilizes a mathematically equivalent sign operation to transform the bimodal distribution into a relatively easy-quantized normal distribution offline. Second, SAM encompasses diverse attention mechanisms (i.e., self-attention and two-way cross-attention), resulting in substantial variations in the post-Softmax distributions. Therefore, we introduce an Adaptive Granularity Quantization for Softmax through searching the optimal power-of-two base, which is hardware-friendly. Extensive experimental results across various vision tasks (instance segmentation, semantic segmentation and object detection), datasets and model variants show the superiority of PTQ4SAM. For example, when quantizing SAM-L to 6-bit, we achieve lossless accuracy for instance segmentation, about 0.5\% drop with theoretical 3.9$\times$ acceleration. The code is available at \url{https://github.com/chengtao-lv/PTQ4SAM}.

Chengtao Lv, Hong Chen, Jinyang Guo, Yifu Ding, Xianglong Liu• 2024

Related benchmarks

TaskDatasetResultRank
Semantic segmentationADE20K (val)
mIoU33.66
3069
Instance SegmentationCOCO 2017 (val)--
1275
Instance SegmentationCOCO (val)--
485
Oriented Object DetectionDOTA v1.0 (test)--
395
Camouflaged Object DetectionCOD10K (test)
S-measure (S_alpha)0.665
306
Object DetectionMS-COCO 2017 (val)
mAP72.9
264
Camouflaged Object DetectionCOD10K
S-measure (S_alpha)0.7257
217
Camouflaged Object DetectionChameleon
S-measure (S_alpha)65.5
207
Instance SegmentationCOCO
mAP57
183
Video Object SegmentationDAVIS
J & F Mean88.51
128
Showing 10 of 18 rows

Other info

Code

Follow for update