Machine learning is still considered "new" or "difficult" so people often respond to posts like this with comments like the one that says "Nothing impressive, existing architectures were used". Writers only get enthusiasm if they raise a benchmark with a brand new approach.
For researchers and people spend a lot of time on ML, I totally get that. There isn't anything new here. In fact, I correctly guessed nearly the entire content of the article from the title alone (transfer learning, using Keras' "Application" models, etc). But why does that matter? It's still a useful article to a lot of people and it's well written.
There is still a dearth of practical advice on how non-specialist developers can solve everyday problems with ML. The truth is that the majority of real projects that real companies need done could be done exactly this way. Using this approach, a developer could get decent results and be done with a project in hours instead of months.
For some reason everyone thinks it's fine to post an article about how to speed up your CRUD website with a Postgres indexes even though it's been done millions of times before, but ML posts get dismissed if they aren't making some kind of research breakthrough. I personally feel like we need a lot more approachable discussion around ML like this in the world. Every dev should feel comfortable getting started with this kind of work if they want to.
Agreed. There's too much of an academic paper culture around ML. It makes the field very unapproachable. It's as if people don't actually use it for anything practical and they only care about the science and math.
It is getting better though. It's great to see people post articles like this and even if it has been done before seeing different approaches and seeing it written in different ways goes a long ways towards educating people.
I'm hoping the JavaScript ML / DL ecosystem expands more since that would greatly accelerate its adoption and allow more developers access to ML.
It's definitely getting better. With fast.ai, and Tensorflow coming with examples and docs to get started, I was able to make a proof of concept in a few hours to solve a real problem for real money.
It feels like we are on the cusp of non-experts being able to get real results and them being able to improve on classifiers for their domain. It almost feels like right before Rails took the world by storm.
Recent tendency in ML/DL academic papers is applying described approach on some public benchmark, and showing improvement over current State of The Art results.
ML/DL is certainly doing better than most but it still has the academic culture. It is increasingly more common to release code on github so the result can be reproduced and people can adapt their research.
While it might not be novelty (and therefor it doesn't seem impressive to some), Transfer Learning is actually highly efficient even for small datasets and should at least be considered before wasting energy and money in building your own convnet. Its a good first step when tackling a new problem in this problem space.
Interesting. I did the same exercise with the pretrained Xception and 2 dense layers at the end and only got 83% accuracy. Thanks for sharing the tweaks you made, that's good to know. I wasn't aware of those techniques.
I'd love to see this network tested on hybrid breeds (but still trained on pure breeds) to see if the probabilistic output would be divided between the two parent breeds.
Nothing impressive, existing architectures were used, InceptionV3, Xception and InceptionResNetV2. Fine-tuning those models resulted in a slightly worse accuracy.
Still it seems like a good write up. Yes, it is not innovative, it provides a good presentation and "here's how I solved this every day, company problem." I think this has a benefit in the Data science. Yes, it glosses over the retrieving and preparing your data section, which I feel is the most work in these problems. However, it gives an idea of how to combine your knowledge and think as a new Data Scientist. So I think it is helpful to some people. As other people said there is a plethora of similar articles in other Software Engineering fields that rise consistently in the front page of Hackernews.
For researchers and people spend a lot of time on ML, I totally get that. There isn't anything new here. In fact, I correctly guessed nearly the entire content of the article from the title alone (transfer learning, using Keras' "Application" models, etc). But why does that matter? It's still a useful article to a lot of people and it's well written.
There is still a dearth of practical advice on how non-specialist developers can solve everyday problems with ML. The truth is that the majority of real projects that real companies need done could be done exactly this way. Using this approach, a developer could get decent results and be done with a project in hours instead of months.
For some reason everyone thinks it's fine to post an article about how to speed up your CRUD website with a Postgres indexes even though it's been done millions of times before, but ML posts get dismissed if they aren't making some kind of research breakthrough. I personally feel like we need a lot more approachable discussion around ML like this in the world. Every dev should feel comfortable getting started with this kind of work if they want to.