This repository contains the official TensorFlow implementation of the MSI-Net (multi-scale information network), as described in the Neural Networks paper Contextual encoder-decoder network for visual saliency prediction (2020) and on arXiv.
Abstract: Predicting salient regions in natural images requires the detection of objects that are present in a scene. To develop robust representations for this challenging task, high-level visual features at multiple spatial scales must be extracted and augmented with contextual information. However, existing models aimed at explaining human fixation maps do not incorporate such a mechanism explicitly. Here we propose an approach based on a convolutional neural network pre-trained on a large-scale image classification task. The architecture forms an encoder-decoder structure and includes a module with multiple convolutional layers at different dilation rates to capture multi-scale features in parallel. Moreover, we combine the resulting representations with global scene information for accurately predicting visual saliency. Our model achieves competitive and consistent results across multiple evaluation metrics on two public saliency benchmarks and we demonstrate the effectiveness of the suggested approach on five datasets and selected examples. Compared to state of the art approaches, the network is based on a lightweight image classification backbone and hence presents a suitable choice for applications with limited computational resources, such as (virtual) robotic systems, to estimate human fixations across complex natural scenes.
Our results are available on the original MIT saliency benchmark and the updated MIT/Tübingen saliency benchmark. The latter are derived from a probabilistic version of our model with metric-specific postprocessing for a fair model comparison.
The code is based on Python v3.6.8 and TensorFlow v1.13.1 and is compatible with both Windows and Linux. We strongly recommend to use TensorFlow with GPU acceleration, especially when training the model. Nevertheless, a slower CPU version is officially supported. To install the required dependencies, use either pip or conda:
pip install -r requirements.txt
conda env create -f requirements.yml
For newer GPUs, you may need to install nvidia-tensorflow to successfully run the scripts (see issues #25 and #26).
The results of our paper can be reproduced by first training the MSI-Net via the following command:
python main.py train
This will start the training procedure for the SALICON dataset with the hyperparameters defined in config.py. If you want to optimize the model for CPU usage, please change the corresponding device value in the configurations file. Optionally, the dataset and download path can be specified via command line arguments:
python main.py train -d DATA -p PATH
Here, the DATA argument must be salicon, mit1003, cat2000, dutomron, pascals, osie, or fiwi. It is required that the model is first trained on the SALICON dataset before fine-tuning it on any of the other ones. By default, the selected saliency dataset will be downloaded to the folder data/ but you can point to a different directory via the PATH argument.
All results are then stored under the folder results/, which contains the training history and model checkpoints. This allows to continue training or perform inference on test instances, as described in the next section.
To test a pre-trained model on image data and produce saliency maps, execute the following command:
python main.py test -d DATA -p PATH
If no checkpoint is available from prior training, it will automatically download our pre-trained model to weights/. The DATA argument defines which network is used and must be salicon, mit1003, cat2000, dutomron, pascals, osie, or fiwi. It will then resize the input images to the dimensions specified in the configurations file. Note that this might lead to excessive image padding depending on the selected dataset.
The PATH argument points to the folder where the test data is stored but can also denote a single image file directly. As for network training, the device value can be changed to CPU in the configurations file. This ensures that the model optimized for CPU will be utilized and hence improves the inference speed. All results are finally stored in the folder results/images/ with the original image dimensions.
A version of the model trained on SALICON was converted to TensorFlow v2.15.0 and made available on Kaggle Hub and HuggingFace Hub. To load the model from one of the platforms, run the following code snippet:
Kaggle Hub
from tensorflow_hub import load
model = load("https://www.kaggle.com/models/alexanderkroner/msi-net/tensorFlow2/salicon/1").signatures["serving_default"]HuggingFace Hub
from huggingface_hub import from_pretrained_keras
model = from_pretrained_keras("alexanderkroner/MSI-Net")For more information on the example usage of the loaded model for predicting saliency from an image, please refer to the guides on Kaggle and HuggingFace. A code notebook for the former can be found here.
A demonstration of saliency prediction in the browser is available here. It computes saliency maps based on the input from a webcam via TensorFlow.js. Since the library uses the machine's hardware, model performance is dependent on your local configuration. The buttons allow you to select the quality, ranging from very low for a version trained on low image resolution with high inference speed, to very high for a version trained on high image resolution with slow inference speed.
Another demonstration to predict where humans fixate on an image can be found on HuggingFace Spaces. Upload an image file, take a snapshot from your webcam, or paste an image from the clipboard to compute the saliency map and download the results.
If you use this code or model in your research, please cite the following paper:
@article{kroner2020contextual,
title={Contextual encoder-decoder network for visual saliency prediction},
author={Kroner, Alexander and Senden, Mario and Driessens, Kurt and Goebel, Rainer},
url={http://www.sciencedirect.com/science/article/pii/S0893608020301660},
doi={https://doi.org/10.1016/j.neunet.2020.05.004},
journal={Neural Networks},
publisher={Elsevier},
year={2020},
volume={129},
pages={261--270},
issn={0893-6080}
}
For questions, bug reports, and suggestions about this work, please create an issue in this repository.