5 ms·

I wonder if this also applies to computer vision? There are certainly problem spaces where heuristics are well established, but many approaches around object de

by joeldo 5y ago

I wonder if this also applies to computer vision? There are certainly problem spaces where heuristics are well established, but many approaches around object detection/segmentation seem much easier/robust to implement with machine learning.

godelski 5y ago
There are some stuff that is more robust but the clarification is hard with classical methods or even small models. Though we're getting better at small models. There's different biases in the models too. But I wouldn't expect classical methods to do well on ImageNet. Though ImageNet has a lot of issues...
crubier 5y ago
Was going to answer this. I very much agree with the article, but deep learning is absolutely a game changer for computer vision. I myself tried several time to “not use ML” for some easy computer vision tasks where traditional CV methods are supposed to work. Well I always end up in situations where they don’t work well without fine parameter tuning, and tuning the parameter for a situation breaks the model in other situations, so you start adding layers of complexity to automatically tune the parameters, but the parameter tuning system also has its own parameters... While a simple neural net is trained easily and is much more robust, saving a lot of time and complexity. Another proof of that is that CV products only started meaningfully entering the market after ML became applicable to CV (after 2015 for complex tasks, or earlier for simpler stuff like MNIST).
- bushbaba 5y ago
  To be fair. Doing that fine parameter tuning and complex layering of heuristics is to some extent creating a “ml model” from hand.
  - crubier 5y ago
    Exactly. This is why usually when you reach that point, a red light turns on in your brain saying “you are just reinventing ML at this point, stop”
- oakfr 5y ago
  The aerospace industry has been using CV successfully for decades, well before ML appeared. So I would temper your last statement a bit. That being said, you are right to say that ML changed the game entirely for CV in industry at large.
  - potatoman22 5y ago
    Same with industrial inspection.
kumarvvr 5y ago
The success of Deep Learning in Computer Vision is fascinating for me. It revolutionized the field in a little over a decade and brought forth new frontiers. I do believe that CV is an area ML will excel well into the next century. Perhaps, we will find a way to chain together ML systems dynamically, overseen by a procedural system that makes real time decisions in understanding its input.

by joeldo 5y ago

godelski 5y ago

There are some stuff that is more robust but the clarification is hard with classical methods or even small models. Though we're getting better at small models. There's different biases in the models too. But I wouldn't expect classical methods to do well on ImageNet. Though ImageNet has a lot of issues...

crubier 5y ago

Was going to answer this. I very much agree with the article, but deep learning is absolutely a game changer for computer vision. I myself tried several time to “not use ML” for some easy computer vision tasks where traditional CV methods are supposed to work. Well I always end up in situations where they don’t work well without fine parameter tuning, and tuning the parameter for a situation breaks the model in other situations, so you start adding layers of complexity to automatically tune the parameters, but the parameter tuning system also has its own parameters... While a simple neural net is trained easily and is much more robust, saving a lot of time and complexity. Another proof of that is that CV products only started meaningfully entering the market after ML became applicable to CV (after 2015 for complex tasks, or earlier for simpler stuff like MNIST).

bushbaba 5y ago
To be fair. Doing that fine parameter tuning and complex layering of heuristics is to some extent creating a “ml model” from hand.
- crubier 5y ago
  Exactly. This is why usually when you reach that point, a red light turns on in your brain saying “you are just reinventing ML at this point, stop”
oakfr 5y ago
The aerospace industry has been using CV successfully for decades, well before ML appeared. So I would temper your last statement a bit. That being said, you are right to say that ML changed the game entirely for CV in industry at large.
- potatoman22 5y ago
  Same with industrial inspection.

kumarvvr 5y ago

The success of Deep Learning in Computer Vision is fascinating for me. It revolutionized the field in a little over a decade and brought forth new frontiers. I do believe that CV is an area ML will excel well into the next century. Perhaps, we will find a way to chain together ML systems dynamically, overseen by a procedural system that makes real time decisions in understanding its input.