Spatial Transformer Nets in TensorFlow/ TensorLayer

Last update: Nov 23, 2022

Overview

MOVED TO HERE

Spatial Transformer Networks

Spatial Transformer Networks (STN) is a dynamic mechanism that produces transformations of input images (or feature maps)including scaling, cropping, rotations, as well as non-rigid deformations. This enables the network to not only select regions of an image that are most relevant (attention), but also to transform those regions to simplify recognition in the following layers.

Video for different transformation click me.

In this repositary, we implemented a STN for 2D Affine Transformation on MNIST dataset. We generated images with size of 40x40 from the original MNIST dataset, and distorted the images by random rotation, shifting, shearing and zoom in/out. The STN was able to learn to automatically apply transformations on distorted images via classification task.

Fig 1：Transformation

Fig 2：Network

Fig 3：Formula

Result

After classification task, the STN is able to transform the distorted image from Fig 4 back to Fig 5.

Fig 4: Input

Fig 5: Output

Comments

Export graph

Hello,

I'm trying to export this model and its weights into a frozen graph. So far I did this in the "save images" part of the training loop:

saver = tf.train.Saver() saver.save(sess, 'my_stn_model_' + str(epoch)) tf.train.write_graph(sess.graph_def, ".", "test.pb", False) #proto

Then I have my pb and the weights. But I am unable to generate a frozen graph from this because I cannot guess the output node in the graph (which is a really complex one it seems).

I attached the pb which I am able to visualize using Netron:

test.zip

Thanks in advance.

opened by hvico 0
Unexpected Output

I'm trying to train the STN. The training dataset which I've provided contains MNIST dataset images which are rotated 90 degree and the testing dataset contains simple MNIST dataset images. The output I'm getting is straight MNIST images. I was expecting the output to contain characters which are rotated 90 degrees. Can you please guide me on how this works?

opened by nabeel3133 4

Spatial Transformer Nets in TensorFlow/ TensorLayer

Related tags

Overview

MOVED TO HERE

Spatial Transformer Networks

Result

You might also like...

Group Activity Recognition with Clustered Spatial Temporal Transformer

VSR-Transformer - This paper proposes a new Transformer for video super-resolution (called VSR-Transformer).

Python implementation of cover trees, near-drop-in replacement for scipy.spatial.kdtree

data/code repository of "C2F-FWN: Coarse-to-Fine Flow Warping Network for Spatial-Temporal Consistent Motion Transfer"

Unofficial implementation of "TTNet: Real-time temporal and spatial video analysis of table tennis" (CVPR 2020)

Official PyTorch implementation of Spatial Dependency Networks.

Spatial Intention Maps for Multi-Agent Mobile Manipulation (ICRA 2021)

The open source code of SA-UNet: Spatial Attention U-Net for Retinal Vessel Segmentation.

Spatial Action Maps for Mobile Manipulation (RSS 2020)

Comments

Export graph

Unexpected Output

Releases(0.0.1)

0.0.1(Jun 26, 2017)

Owner

Hao

SAGE: Sensitivity-guided Adaptive Learning Rate for Transformers

Using Streamlit to host a multi-page tool with model specs and classification metrics, while also accepting user input values for prediction.

Code for our TKDE paper "Understanding WeChat User Preferences and “Wow” Diffusion"

KeypointDeformer: Unsupervised 3D Keypoint Discovery for Shape Control

DvD-TD3: Diversity via Determinants for TD3 version

Official implementation for the paper: Multi-label Classification with Partial Annotations using Class-aware Selective Loss

Official code repository for A Simple Long-Tailed Rocognition Baseline via Vision-Language Model.

[Arxiv preprint] Causality-inspired Single-source Domain Generalization for Medical Image Segmentation (code&data-processing pipeline)

PyTorch implementation of SmoothGrad: removing noise by adding noise.

LRBoost is a scikit-learn compatible approach to performing linear residual based stacking/boosting.

Localized representation learning from Vision and Text (LoVT)

PASSL包含 SimCLR，MoCo，BYOL，CLIP等基于对比学习的图像自监督算法以及 Vision-Transformer，Swin-Transformer，BEiT，CVT，T2T，MLP_Mixer等视觉Transformer算法

Snscrape-jsonl-urls-extractor - Extracts urls from jsonl produced by snscrape

Coded illumination for improved lensless imaging

Pytorch implementation of Compressive Transformers, from Deepmind

An implementation demo of the ICLR 2021 paper Neural Attention Distillation: Erasing Backdoor Triggers from Deep Neural Networks in PyTorch.

Improving Query Representations for DenseRetrieval with Pseudo Relevance Feedback:A Reproducibility Study.

Motion planning environment for Sampling-based Planners

PyTorch Autoencoders - Implementing a Variational Autoencoder (VAE) Series in Pytorch.

ML models implementation practice