Roboflow Reviews & Product Details

I like how easy it is to use Roboflow's interface. It gives me so many options for training computer vision models and downloading data for further use. It is also super easy to split work up between my team and I so we can quickly finish tasks and share the workload. I also like the annotation options which allow me to either draw quick squares or draw out a good polygon if needed. Review collected by and hosted on G2.com.

Sometimes, but not always, it can be difficult to remove errors when annotating images. You have to scroll through to find the exact annotation you want to delete, which in my case can take a lot of time due to the fact that I typically am annotating 150+ times per image. S

Sometimes implementing the full project life cycle can be difficult. You have to have a strong background in how these types of programs work in order to get the best bang for your buck using this program. Review collected by and hosted on G2.com.

What I like the most about Roboflow is the simplicity and ease of labeling, training, testing, and deploying AI models. I really appreciate being able to upload already-labeled images and adjust the annotations directly on the webpage. The platform is very user-friendly for newcomers and has always been quite intuitive for tasks like data augmentation, for example. I use it almost daily whenever I need to improve my datasets. It’s very easy to integrate into your AI workflow or pipeline for training. Customer support has always been good, at least in the few times I’ve had to reach out to them. Review collected by and hosted on G2.com.

The only drawback I see is that in order to have a private image repository, you need a subscription; otherwise, everyone can see your repository and dataset. Also, sometimes when uploading datasets, certain parts remain permanently on the images and can’t be removed, though this is something that rarely happens. Review collected by and hosted on G2.com.

Roboflow makes it incredibly easy to manage datasets for computer vision tasks. Uploading, annotating, and preprocessing images is straightforward, even for large datasets. I really like the built-in augmentation options, which save a lot of time when preparing data for training. The integration with popular machine learning frameworks (like YOLO, TensorFlow, and PyTorch) is seamless, so exporting datasets into the right format is hassle-free. The interface is clean and beginner-friendly, but also powerful enough for advanced users. Review collected by and hosted on G2.com.

Some of the more advanced features are locked behind paid plans, which can be limiting for students or small projects. The free tier is helpful but has restrictions on dataset size and usage. Review collected by and hosted on G2.com.

Roboflow makes dataset management and model training straightforward. I really like how easy it is to annotate, preprocess, and version datasets without having to build a whole custom pipeline. The integrations with popular ML frameworks save a ton of time, and the UI is clean and intuitive. We use this within our business daily and the support is always available for any issue we come across. Review collected by and hosted on G2.com.

Some of the more advanced features are locked behind higher pricing tiers, which can be limiting for smaller projects or hobby use. It would also be nice to have a bit more flexibility in automation pipelines without needing to step outside the platform. Review collected by and hosted on G2.com.

I like Roboflow in different ways. But the best thing about Roboflow is the user friendly interface. I can have data, label data, segment data, augment data and train data in a single place. I like the smart polygon labeling. It can learn from the previous labeling. I can augment data in several ways and can make the variation of the size of the data. The data preprocessing tool is very helpful to make variations in data. Overall from my 6 to 7 research i used roboflow as primary data repository and i mentioned it in my every published research papers. Easy implementation tool like Roboflow is what we need. Review collected by and hosted on G2.com.

There are some points Roboflow can be improved. One of them is the smart polygon segmentation. It's good but not the best. sometimes it needs manual segmentations as the smart polygon struggles to identify objects from previous labeling. It should be more precise. Review collected by and hosted on G2.com.

The speed with which I can annotate, train, and deploy models is incredible. Integration with different model architectures and the ability to deploy them in different environments (web, mobile, etc.) make it a complete and highly versatile solution. Review collected by and hosted on G2.com.

While the free version is very useful, the limitations on the number of images and advanced features in the paid plans can be a bit restrictive for personal or small-scale projects. Review collected by and hosted on G2.com.

Roboflow is a platform designed for working on Computer Vision projects. From the platform, you can upload images or videos, label them, and start generating the first models for detection, classification, or pose estimation, among others. The API is very easy to use and also allows you to label datasets and train them on platforms outside of Roboflow, in case you have dedicated hardware. It is a very useful and essential platform to accelerate developments and facilitate the task of vision engineers. Review collected by and hosted on G2.com.

Although it is very user-friendly, it requires a minimum knowledge of computer vision, especially to evaluate the training metrics. The auto-training, although very easy to use, does not always return optimal results, so I still recommend tuning the parameters on your own based on the dataset and your needs. Review collected by and hosted on G2.com.

It is so easy to train vision models with the UI. You can prototype a model in minutes to validate ideas and, if it works, you can refine it to production-ready on the same platform. It's easy to switch models, but that is not needed with Roboflow's RF-DETR since it performs so well. Review collected by and hosted on G2.com.

Sometime the annotation UI is a little buggy but a refresh always fixes the issue. Review collected by and hosted on G2.com.

It has been extremely easy to access various computer vision strategies without having to learn the intricacies of each. For example, I was able to train a keypoint detection model and then run it against a video of vehicles in under a day. Normally there'd be debugging for days trying to find where code has gone wrong, but they have a plug-n-play option with "Workflows" to use established code chunks and keep rolling forward. And then on top of that, they provide a way to access that same workflow via Python and expand on it as needed for your custom use-case. Review collected by and hosted on G2.com.

The downside is that there are so many ways to apply computer vision, sometimes you stumble on one that they have not formally implemented so then you have to go the manual route. Kind of the case with any complex software platform like this. Although they have also been receptive in taking feedback from individuals for feature requests and so the service is expanding daily. But overall, they seem to have what I need 99% of the time so I always go there first and usually get what I'm looking for. Review collected by and hosted on G2.com.

We used several datasets in our recent paper on out-of-stock detection from a robotic platform [Deep learning based empty shelf detection based on autonomous mobile robot https://www.sciencedirect.com/science/article/pii/S1077314226000640].

They were very easy to find and use. Review collected by and hosted on G2.com.

More persons should upload their data :-) Review collected by and hosted on G2.com.