How to build deep learning image dataset quickly?
Answer: crawl images from yandex reverse image search results, for clean and meaningful images.
I wrote my own crawler using selenium Inspired by @PyImageSearch
Also i used PIL to verify images dimensions on the go instead.
@roshanonline The reason why we succeed in IT industry is mostly because it does not require a lot of investment and dependence on others. However, for other areas, it is still the same issue.
@roshanonline The skill and learning phase and access to indians are broken, I see on yt how people are just building what ever they want and have the skills and tools, but for most of us we don't have that privilege.
3/3
A year ago:
Weekdays = coding
Weekends = cleanup
Today:
Weekends = agents coding
Weekdays = cleanup
Not ideal. More debt. More tradeoffs.
But we're shipping faster than ever, and for non-critical systems I think it's worth it.
1/3
"if it works, don't touch it."
AI agents changed that for me.
They're generating feature updates so quickly that I can't realistically review every line anymore, even if I wanted to.
2/3
AlgoFly AI's architecture took me 2 years to build.
Most of the time Codex follows it surprisingly well.
Sometimes it doesn't.
I can usually spot the divergence, but I still want the feature shipped ASAP.
That's the tradeoff.
@PINTO03091 you can try my platform https://t.co/IIiI7Q68e4 it's similar to roboflow but with better ui and we are giving free on prem deployments for first few users