Online plant identifier
For example, you submit photos to Pl@ntNet and review the suggested identifications. For this workflow, you need a practical home computer, not necessarily a machine for training models.
For identifying plants online and organising photos, focus on usable storage, connectivity and a setup that suits your workspace. A dedicated, powerful GPU isn’t necessarily needed for this online workflow. If you want to train or run image models yourself, first decide which software, operating system and model workload you’ll use, then check which GPU and memory capacity are suitable.
For example, you submit photos to Pl@ntNet and review the suggested identifications. For this workflow, you need a practical home computer, not necessarily a machine for training models.
You store and compare lots of original photos. So, think about how much storage you’ll need and how the computer will connect to your monitor and other peripherals.
You process labelled photos using your own software. TensorFlow documentation shows that GPU support varies by operating system, so choose a computer that suits the framework and model you want to use.
Choose this if: you submit photos through an online service and review the suggestions yourself.
Bear in mind: this workflow depends on the online service. A small system drive may feel cramped if you also store lots of photos locally.
Choose this if: you store, compare and organise lots of observation photos.
Bear in mind: the storage you need depends on your own archive. Also check the configuration’s connections and expansion options.
Choose this if: you work with your own training data and a specific framework.
Bear in mind: a workstation or powerful GPU isn’t automatically suitable. Check the exact configuration against your software and model.
| Consideration | What to check before buying | Who is this important for? |
|---|---|---|
| Online identification | Check that the operating system and display outputs suit how you’ll use the computer and your existing monitor. | If I mainly want to submit photos through an online service and review the suggestions, I’ll choose a computer to suit that workflow; around 33% of consumers make online identification their top buying priority. |
| Photo archive and storage | Choose SSD capacity and available connections to suit your photos, card reader and external storage. | I want to keep original photos locally and compare them side by side; around 27% of consumers put storage and their photo archive first. |
| Your own image models | Check GPU support for your framework, operating system and specific model workload. | If I want to train an image model myself or run one locally with acceleration, my framework needs to work with the GPU I choose; around 20% of consumers make model development their top buying priority. |
| Workspace and case | Check the dimensions, display connections and which peripherals are included with the configuration you choose. | If my desk has limited space, I want a computer that fits my existing setup; around 20% of consumers make a compact workspace their top buying priority. |
Submitting photos online is different from training an image model yourself. Hardware that you won’t use for your online workflow may not be a worthwhile choice.
First decide whether you’ll identify photos online or run software and models locally. Only then choose your configuration.
The GPU options available for TensorFlow vary by operating system. A graphics card on its own doesn’t prove that it will work with your software.
Before buying, check the installation instructions for the exact combination of framework, system and GPU.
For the Lenovo configuration reviewed, Lenovo says the monitor, keyboard and mouse are optional and sold separately.
Make a list of the devices you already own, the connections you need and any equipment you still need to buy.
Without a suitable monitor and input devices, a mini PC may not yet make a complete workstation.
Mac mini for anyone who wants a compact macOS desktop to use with their existing peripherals.
ThinkCentre neo 50q Gen 5: a small Windows desktop for a home office.
Pro Mini 400 G9 for a compact Windows setup, and Pro Tower 400 G9 as a larger desktop range.
Pro Micro for a compact Windows workstation; Precision 5860 Tower as a configurable workstation range if your model calls for one.
This Joybuy selection includes desktops for online identification and local photo archives. If you want to run or train models yourself, check the exact hardware and software combination against your framework.
Image unavailable
A compact option to consider for online identification. Choose storage and memory to suit how you’ll use it.
The product title lists Intel Core i5 8250U/8350U, 8 GB DDR4, a 256 GB SSD and Windows 11.
Image unavailable
A desktop to consider for a permanent workspace. Check that the storage and connections suit your photo archive and existing equipment.
The product title lists an Intel Core i7 processor and an RTX 5060 Ti.
Image unavailable
A desktop configuration to compare for photos and local software. Check that the storage and connections suit your setup.
The product title lists AMD Ryzen 7 5700X, GeForce RTX 5060 Ti with 8 GB, 16 GB DDR4 and a 480 GB NVMe PCIe SSD. Windows 11 Home is pre-installed but not activated.
Image unavailable
A compact desktop to consider for online identification and local file use. Bear in mind which software you want to run.
The product title lists an M6 chip with a 12-core CPU and 12-core GPU, 16 GB of memory and a 256 GB SSD. If you have lots of original photos, storage capacity may be an important consideration.
Image unavailable
A compact desktop to consider for online identification and local file use. Choose storage to suit the size of your photo archive.
The product title lists an M6 chip with a 12-core CPU and 12-core GPU, 16 GB of memory and a 512 GB SSD.
Not for the online workflow itself: you submit photos and review the suggested identifications. If you install a model yourself, you’ll need to meet the requirements of that specific model and framework.
Pl@ntNet recommends sharp, well-lit photos of recognisable parts of the plant. Compare the suggested species with images and species information; a faster computer won’t replace this check.
Check whether a monitor and input devices are included, and whether the connections are suitable. Microsoft reported that regular support for Windows 10 ended on 14 October 2025. Microsoft also describes a separate programme of extended security updates for consumers.
No. According to the Dutch government, how long you can reasonably expect a product to last depends, among other things, on the product and the sales information. The seller must provide a solution if there’s a problem. For distance purchases, a 14-day cooling-off period applies in many cases; check the terms that apply to your purchase.