Proprietary/Spinenet 미리보기

SpineNet is an image object detection model generated using neural architecture search. Du et al (2020) proposed it in the paper "SpineNet: Learning Scale-Permuted Backbone for Recognition and Localization." A Google internal dataset is used to pretrain the model. The pretrained checkpoint will be loaded as the initial checkpoint. The derived models can be used in commercial products but the weights can't be exported.

Convolutional neural networks typically encode an input image into a series of intermediate features with decreasing resolutions. While this structure is suited to classification tasks, it does not perform well for tasks requiring simultaneous recognition and localization (e.g., object detection). The encoder-decoder architectures are proposed to resolve this by applying a decoder network onto a backbone model designed for classification tasks. Du et al argue that encoder-decoder architecture is ineffective in generating strong multi-scale features because of the scale-decreased backbone. SpineNet is a backbone with scale-permuted intermediate features and cross-scale connections that is learned on an object detection task by neural architecture search.

Using similar building blocks, SpineNet models outperform ResNet-FPN models by ~3% AP at various scales while using 10-20% fewer FLOPs. In particular, SpineNet-190 achieves 52.5% AP with a MaskR-CNN detector and achieves 52.1% AP with a RetinaNet detector on COCO for a single model without test-time augmentation, significantly outperforming prior state-of-the-art detectors. SpineNet can transfer to classification tasks, achieving 5% top-1 accuracy improvement on a challenging iNaturalist fine-grained dataset.

This model is implemented in the TPU Object Detection and Segmentation Model Zoo repository on GitHub.

• Image object detection: identifies objects in images with bounding boxes. Models can be trained to detect common objects like vehicles, furniture, packaged goods, etc.

Get started

This model can be used in a notebook. Click Open notebook to use the model in Colab.

Dataset and training

This model checkpoint was pretrained on Google's proprietary dataset. Training images are resized/rescaled to the same resolution (1024x1024) and normalized across the RGB channels with mean (0.5, 0.5, 0.5) and standard deviation (0.5, 0.5, 0.5). For more details, see the notebook.

Model output

Given an image, the model will output a vector of confidence scores for each support class (label) the model identifies.

Model output

Given an image, the model will output x-axis aligned bounding boxes, and corresponding object classes with confidence scores.

Versions

Links

• SpineNet: Learning Scale-Permuted Backbone for Recognition and Localization by Du et al (2019)
• TPU Object Detection and Segmentation Model Zoo on GitHub
• TPU Object Detection and Segmentation Framework on GitHub
• SpineNet: A Novel Architecture for Object Detection Discovered with Neural Architecture Search