Rowel Atienza 🇵🇭

@Em_alej HiFiGAN

@jacobe Amazing work, it generates spectrograms, but what is it using as a vocoder? I just found this repo while looking for something more efficient than Tacotron.

Most ARM chips can't run decent AI models. Introducing EfficientSpeech, a 266k-param TTS model. Low cost ARM chips like in RPi4 can generate 104sec of speech mel spec in 1sec. Here's an AI-generated video w/ voice from EfficientSpeech. Info: github.com/roatienza/effi… #ICASSP2023

We are lucky to have this paper accepted at ECCV 2022! More details later ... #ECCV2022

@arXiv_Daily Simple yet effective idea: Remove inefficient top-most layers & replace them with an efficient head. For VWW, param count reduced by 93% with only 0.65% accuracy decrease. Counterintuitively, the quantized pruned net increased its accuracy on ARM Cortex M0.

Depth Pruning with Auxiliary Networks for TinyML deepai.org/publication/de… by Josen Daniel De Leon and @jacobe #NeuralNetwork #ComputerScience

.@Gradio & @huggingface for Rapid Deep Learning App Development by @jacobe link: @rowel/gradio-hugging-face-for-rapid-deep-learning-app-development-709a78e7ccc0" target="_blank" rel="nofollow noopener">medium.com/@rowel/gradio-…

@ak92501 @Gradio @huggingface Thanks. Let's keep on building better infrastructure and tooling to make AI more accessible.

"The UP National Engineering Center Analytics and Data Science Certifications announced the development on Wednesday." gmanetwork.com/news/scitech/t… via @gmanews

Idea: If data augmentation improves model generalization, why not use it to generate 2 new inputs and force the representations to agree. Result: Additional model performance improvement. Comparison: Unlike Label Smoothing, the performance of our method, AgMax, is consistent.

I am excited to share my latest work: 8-bit optimizers – a replacement for regular optimizers. Faster 🚀, 75% less memory 🪶, same performance📈, no hyperparam tuning needed 🔢. 🧵/n Paper: arxiv.org/abs/2110.02861 Library: github.com/facebookresear… Video: youtube.com/watch?v=IxrlHA…

We’re introducing GSLM, the first language model that breaks free completely of the dependence on text for training. This “textless NLP” approach learns to generate expressive speech using only raw audio recordings as input. Learn more and get the code: ai.facebook.com/blog/textless-…

@arXiv_Daily Data Augmentation for STR will be presented at ICCV 2021 Workshop on Interactive Labeling and Data Augmentation for Vision. GitHub: github.com/roatienza/stra…

Data Augmentation for Scene Text Recognition deepai.org/publication/da… by @jacobe #ComputerVision #ComputerVision

ODSC APAC Virtual Conference Speaker: hubs.ly/H0V892b0 #ODSCAPAC #DataScience #DeepLearning @jacobe @upsystem

My 1st github project that reached the 1k milestone. @PacktPub github.com/roatienza

@arXiv_Daily It took us more than a year building, collecting, annotating, validating and benchmarking this dataset. Dataset: github.com/upeee/GOO-GAZE… To appear at #CVPR2021 Workshop: gazeworkshop.github.io/2021/

GOO: A Dataset for Gaze Object Prediction in Retail Environments deepai.org/publication/go… by Henri Tomas et al. including @jacobe #Estimator #Statistics

@Deep__AI Vision Transformer (ViT) for reading real-world text images. My paper will be presented at #ICDAR2021 icdar2021.org.

Vision Transformer for Fast and Efficient Scene Text Recognition deepai.org/publication/vi… by @jacobe #ComputerVision #PatternRecognition

Yesterday, my former grad student Daryl gave a talk at Sony CSL Paris about his thesis on Next View Policy for 3D Reconstruction. Youtube: youtu.be/KdyDj3bjU0I