Lossy Image Compression with Compressive Autoencoders

Authors: Lucas Theis, Wenzhe Shi, Andrew Cunningham, Ferenc Huszár

ICLR 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We compared our method to JPEG (Wallace, 1991), JPEG 2000 (Skodras et al., 2001), and the RNN-based method of (Toderici et al., 2016b)4. We evaluated the different methods in terms of PSNR, SSIM (Wang et al., 2004a), and multiscale SSIM (MS-SSIM; Wang et al., 2004b).
Researcher Affiliation Industry Lucas Theis, Wenzhe Shi, Andrew Cunningham& Ferenc Husz ar Twitter London, UK {ltheis,wshi,acunningham,fhuszar}@twitter.com
Pseudocode No The paper describes the architecture and method in text and diagrams, but does not include pseudocode or algorithm blocks.
Open Source Code No The paper does not provide a statement or link to their own open-source code for the methodology described. It only references third-party code used or adapted.
Open Datasets Yes For testing, we use the commonly used Kodak Photo CD dataset of 24 uncompressed 768 512 pixel images3. 3http://r0k.us/graphics/kodak/
Dataset Splits No Hyperparameters affecting network architecture and training were evaluated on a small set of held-out Flickr images.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments, only general statements about the computational efficiency of their network.
Software Dependencies Yes All networks were implemented in Python using Theano (2016) and Lasagne (Dieleman et al., 2015).
Experiment Setup Yes All models were trained using Adam (Kingma & Ba, 2015) applied to batches of 32 images 128 128 pixels in size. ... the learning rate is reduced from an initial value of 10 4 to 10 5. Training was performed for up to 106 updates... Here we used an initial learning rate of 10 3 and continuously decreased it by a factor of τ κ/(τ + t)κ, where t is the current number of updates performed, κ = .8, and τ = 1000. Scales were optimized for 10,000 iterations.