Review on PaddlePaddle by Ron Long

I liked that it is easy to learn (it has an excellent documentation) and can be used for many different tasks from image classification or object detection in videos to text recognition. The model could take some time before giving you good results but once they are there, they're great! It works well for most of my use cases so far - even though sometimes training takes longer than expected due to memory limitations.- Sometimes models give bad outputs when being presented with similar images as input like "cat" classifier gives me mostly dogs instead.- No GPU support for now which limits its usage compared to other frameworks such as TensorFlow/PyTorch. But this might change soon since developers have been working hard towards making them compatible. A common problem solved using these tools was identifying products at point of sale. Using their facial identification feature we were able to achieve over 90% accuracy rate while processing around 10,000 faces per second! I like that it is very easy to use. It has a lot of great features and you can easily integrate with other tools. The documentation could be better. There are some things I don't understand fully. But overall it works well for my needs. I am using it to do speech recognition and text generation. It does this really well. It is one of the best options out there for these types of tasks.