manneshiva

This is a step-by-step tutorial on how to use tensorflow.js to run an object detection web application for your webcam feed in real-time. The model used is COCO-SSD-mobilenet_v2.

Interesting, wonder what their use case was?

How fast are the algorithms? does anyone of them work in real time?

A news agency in China uses an AI news anchor.

The speed at which China seems to be adopting and implementing the latest deep learning technology into everyday life is truly astounding. There is no doubt it is going to be one of the top countries contending for the world leader in AI in the next decade or so.

I have seen similar results using Gunner Farneback's algorithm.

Highlights of this technique:

Simple: One-sentence method summary: use keypoint detection technic to detect the bounding box center point and regress to all other object properties like bounding box size, 3d information, and pose.

Versatile: The same framework works for object detection, 3d bounding box estimation, and multi-person pose estimation with minor modification.

Fast: The whole process in a single network feedforward. No NMS post processing is needed. Our DLA-34 model runs at 52 FPS with 37.4 COCO AP.

Strong: Our best single model achieves 45.1AP on COCO test-dev.

Easy to use: We provide user friendly testing API and webcam demos.

Even Facebook uses Bayesian Optimization for tuning the parameters of some of its online systems (like its internal ranking system): https://research.fb.com/efficient-tuning-of-online-systems-u...

Sure, it does, it's not trivial though. Tedious to implement it yourself. You could use python libraries as "scikit-optimize" which has an implementation of parallel Bayesian optimization (based on Gaussian process), have a look at this: https://scikit-optimize.github.io/notebooks/bayesian-optimiz...

Manually searching is time taking since you need to wait for the results from each experiment. This becomes impossible when the number of hyperparameters is more than 8-10 and you will probably end up only tuning a few of them that you think are relevant. You'd also need a lot of experience in tuning hyperparameters else your tuning is as good as random.

Given these disadvantages of manual tuning, "Bayesian Optimization" seems like the most promising technique, it needs a lot less "choose->train->evals" loops as it uses the information from previous runs to select the next set of hyperparameters (similar to what humans would do).

It just amazes me how the attention to detail in a video game could help reconstruct a medieval Catholic cathedral!

Training deep learning models can be tough. They don't work without the right hyperparameters. This interactive blog gives an explanation of the algorithms that can automate the hyperparameter search and has all the code you need to try it out for yourself.

If you've been conducting manual quality checks at your manufacturing company, you've probably been over-paying for low productivity and poor quality output. The link explains why AI-powered visual inspection is the future of manufacturing.

Training deep learning models can be tough. They don't work without the right hyperparameters. The interactive blog gives an explanation of the algorithms that can automate the hyperparameter search and has all the code you need to try it out for yourself.

Sure, all the papers are one of the best in their respective fields, no doubt about that.

Regarding your paper, do you think similar fine timing techniques can be applied to other NLP tasks like NER? Did you get a chance to try on any other non classification task?

Btw, kudos to you guys for such a practically useful paper.

My 2 favorites are 1. BERT - easy to understand, SOTA results on most NLP tasks, code and pretrained models readily available. What more can you ask! 2. ULMFit - transfer learning for NLP, have been waiting for something like this since a long time.

Both these mean getting better results/models for most NLP tasks with lesser training data, effort and time.

Here is another link explaining these topics in a little more detail (a must read if you are in the NLP field) -- https://jalammar.github.io/illustrated-bert/

Here are a few more who have tried it or use deep learning for SAR: https://www.planckaero.com/maritimedrone/ https://flytbase.com/flytstories/search-and-rescue-drone/ https://news.developer.nvidia.com/autonomous-search-and-resc...

Data for training such models for practical use are difficult to get. Often, one has to train a model on a different but related dataset that contains your object of interest and ensure that your model generalizes well. These models are highly sensitive to the data that they are trained on, so it is best if you can gather data which is closer to the actual environment (in which you want to use the model).

just curious, how do you intend to use this? I mean, what problem does exploring each layer in a Docker solve for you?