Here’s an interesting computer vision project that should make recycling easier.

Trash Classifier Example Project The Trash Classifier project, affectionately known as “Where does it go?!”, is designed to make throwing things away faster and more reliable. The Trash Classifier project uses a Machine Learning model trained in Lobe to identify whether an object goes in the garbage, recycling, compost, […]

We live in an amazing time where even beginners can take on project that  were the sole domains of expert researchers a decade ago or were straight up science fiction.

Here’s a great write up on the differences between PyTorch and TensorFlow when it come to object detection.

As I already was an experienced data scientist and had developed a few productive machine learning software components in the past years, I thought it should be a fairly easy job to get online, find a pre-trained CNN and train it further on an unseen data set to enable it to detect some previously unknown objects. Up until about a a year ago I had mostly worked with tree-based models, a lot of scikit-learn as well xgboost and of course tons and tons of pandas. As with many AI tasks, my first approach with this one turned out to be a classic version of a: “Not so fast!“ There are a few stepping stones on the way that you have to know of, that however not many articles and blogs seem to mention. After having spent many hours on this topic and having read a lot of TensorFlow source code I know the answers to questions like:

Computer have wide applications across industries for quality control.

For instance, the majority of all medical data is image-based: The assessment of X-rays and scans is crucial for the right diagnosis and, thus, for the right treatment.

Public health depends on the accurate interpretation of every single image, and many physicians are obliged to choose between longer working hours or doing less detailed and precise medical image analysis.

Right now, medical staffers around the world are stretched thin, this is where AI can come into play.

Artificial intelligence in healthcare speeds up the process of medical image analysis and makes it more accurate and stress-free for medical personnel. Using artificial intelligence, it is possible to detect rare diseases, such as Noonan syndrome, or identify viruses and bacteria by analyzing Petri dish images. Computer vision and machine learning for medical image analysis are becoming as vital as an experienced lab worker with modern equipment.

MIT Introduction to Deep Learning 6.S191: Lecture 6 with Ava Soleimany.

Subscribe to stay up to date with new deep learning lectures at MIT, or follow us @MITDeepLearning on Twitter and Instagram to stay fully-connected!!

Lecture Outline

  • 0:00 – Introduction
  • 0:58 – Course logistics
  • 3:59 – Upcoming guest lectures
  • 5:35 – Deep learning and expressivity of NNs
  • 10:02 – Generalization of deep models
  • 14:14 – Adversarial attacks
  • 17:00 – Limitations summary
  • 18:18 – Structure in deep learning
  • 22:53 – Uncertainty & bayesian deep learning
  • 28:09 – Deep evidential regression
  • 33:08 – AutoML
  • 36:43 – Conclusion

If you’ve ever attended one of my neural network talks, you know that I point out that what neural networks learn is not what you think they actually learn.

As we come to rely on AI to make increasingly more important decisions, we may want to pause and realize that our training data could be used as a vector for bad actors.

The papers, titled “Adversarial Preprocessing: Understanding and Preventing Image-Scaling Attacks in Machine Learning” [PDF] and “Backdooring and Poisoning Neural Networks with Image-Scaling Attacks [PDF],” explore how the preprocessing phase involved in machine learning presents an opportunity to fiddle with neural network training in a way that isn’t easily detected. The idea being: secretly poison the training data so that the software later makes bad decisions and predictions.

The wide array of options for body scanning have improved patient outcomes, but the process of image interpretation is still labor intensive.

Here’s an interesting article on applying AI to the problem.

Malignant brain tumors are one of the most deadly forms of cancer, partially due to the dreadful diagnosis, but also because of the direct consequences on decreased cognitive function and lasting adverse impact on the quality of life of the patient.

Machine Learning with Phil ponders the question: “is it better to specialize or generalize in artificial intelligence and deep learning?”

The answer depends on your career aspirations. Do you want to be a deep learning research professor?

Do you want to go to work for Google, Facebook, or other global mega corporations?

Or do you want to be your own unicorn start up founder?

Each has their own specialization requirements that Phil breaks down in this video.

Siraj Raval explores why does a computer algorithm classify an image the way that it does? This is a question that is critical when it comes to AI applied to diagnostics, driving, or any other form of critical decision making.

In this video, he raises awareness around one technique in particular that I found called “Grad-Cam” or Gradient Class Activation Mappings.