How to choose a Laptop for Computer Vision with Deep Learning (in 2022)

Choosing a laptop is currently the best choice for computer vision and deep learning. In fact, due to the shortage of microchips in manufacturing and mining, the prices of video cards are very high and the laptop is a good alternative.

We will see how to choose a laptop, usable in computer vision with good results, based on the main characteristics.

Choose laptop for computervision

To choose the right laptop the main component to consider is the graphics card. In this, the reference brand is Nvidia because most of the libraries are compatible with this graphics card.

If you are on a low budget let’s see what are the minimum features to consider when choosing your laptop

GPU for laptop

6 GB is the minimum memory needed to use the computer with deep learning. In fact, with Tensorflow or PyTorch, it would fail to work and you would get errors.

CPU for laptop

For the CPU there are no strict requirements but obviously higher is better. Minimal advice

Intel i5 or greater
AMD ryzen 5 or greater

Ram for laptop

8 GB it is the minimum of memory required even if it does not allow fluid use. If you are on a low budget always consider expanding the memory, so buy a laptop with at least 2 RAM slots.

How to choose the right laptop

As we have seen above, the most important thing when you want to buy a laptop is the video card. How do we know when it’s right for us? The simplest thing to do is to look at the video card name and check it on the manufacturer’s website.

For example, if there was written Nvidia Geforce GTX 1650 going to check on the https://www.nvidia.com/en-eu/geforce/graphics-cards/gtx-1650/ site you would find this table:

Circled in red there is the value to be taken into consideration, it says << Frame buffer 4GB >>. In accordance with what we have established on video cards, it must have a minimum memory of 6GB so the computer with this specific video card is not good.

Laptop limitations

Using the laptop with a video card is a good solution but obviously to several limitations that do not make it perfect. For example, the most powerful laptop video card has a maximum of 8GB of memory and this is because they are designed for gaming.

You can certainly work with object detection, instant segmentation with mask R-CNN, instant segmentation but there are limitations in training, for example, the maximum image size.

In conclusion, this can be a good solution if you are starting out, for educational use or if you are on a low budget but it is a terrible solution for professional use. In this case, I would recommend you to rent a server for training or build it yourself