Roboflow Reviews & Product Details

Roboflow provides an end-to-end pipeline for computer vision research, from dataset annotation and versioning to augmentation and model export. The interface is intuitive, and the dataset management features (splits, class balance visualization, augmentation control, and format conversion) significantly reduce experimental overhead. For academic research, the ability to quickly iterate on datasets, reproduce experiments, and export to multiple frameworks (YOLO, COCO, TensorFlow, ONNX, etc.) is extremely valuable. Review collected by and hosted on G2.com.

For very large-scale datasets, cloud processing and export can sometimes be time-consuming, and advanced customization of augmentation pipelines is slightly more constrained compared to fully custom local scripts. Additionally, some advanced features are locked behind paid tiers, which may be a limitation for students with limited funding. Review collected by and hosted on G2.com.

Easy to user interface, great free datasets, loads fast. Quite affordable. Customer support is bad, got no response on trying to upgrade to Enterprise - which has most of the needed features for teams. Using it for over a month. AI auto label works well, which is the main selling point i guess. Review collected by and hosted on G2.com.

The RAM it consumes, we loaded 20 000 images into labelling and it takes 2GB. Just one tab of their site. Looks like they preload all images into frontend.

There is no way to see what other labelers labelled. Requested twice to upgrade to Enterprise but got no response over time.

On object detection, it is way too easy to misspell the class name and by mistake create new class. There must be some way of validating or modal showing up to let the user know that they are creating new class. Thats during labelling. E.g. class name is ar_5 , i typed ar 5 and clicked enter by mistake - done, object got new class. Review collected by and hosted on G2.com.

Roboflow distinguishes itself as a leading end-to-end computer vision platform by streamlining the often disjointed process of object detection and instance segmentation into a cohesive, unified workflow. One of the most notable benefits is the annotation experience, particularly the "Smart Polygon" feature powered by the Segment Anything Model (SAM). This tool transforms the usually painstaking, pixel-level clicking required for segmentation masks into a fast, one-click operation. In addition to this convenience, the platform provides significant strategic value through its powerful preprocessing and augmentation tools, which let you expand your dataset and enhance model robustness by simulating conditions like noise or rotation—without the need to gather new images. Coupled with "Label Assist" for automated pre-labeling and the capability to export datasets in more than 30 widely used formats (including YOLO, COCO, and TFRecord), Roboflow effectively eliminates the technical barriers between data preparation and training. This allows you to dedicate your efforts to refining model performance instead of dealing with file conversions. Review collected by and hosted on G2.com.

What I find frustrating about Roboflow is that its AI features act more like assistants than fully autonomous agents, especially regarding the demanding tasks of annotation and Quality Assurance (QA). When I upload a dataset, I expect the AI to handle image labeling from start to finish. However, I still have to go through each image myself to apply or confirm labels, which makes the process feel unnecessarily manual. This issue becomes even more pronounced with instance segmentation, where precise polygon annotations are essential. The platform does not provide an automated QA layer to guarantee the accuracy of these annotations, so I am left with the responsibility of checking every mask for correctness. I wish there were an AI 'supervisor' to automatically validate the quality and consistency of the data, rather than leaving all the verification work to me. Review collected by and hosted on G2.com.

What I like best about Roboflow is how seamlessly it handles every step of dataset preparation — from uploading and annotating images to preprocessing, augmentation, and exporting for training. The platform is extremely user-friendly, and it saves an incredible amount of time when working with large datasets. The automated features, version control, and YOLO export options made it easy to manage our BharatPotHole dataset of 7,000+ frames efficiently. For our iWatchRoad pothole detection research, Roboflow helped maintain consistent labeling quality, supported collaboration, and simplified the entire workflow. It’s one of the few tools that genuinely makes computer vision dataset management effortless. Review collected by and hosted on G2.com.

Honestly, there’s not much to dislike. Occasionally, large uploads can take a bit longer, especially when dealing with thousands of high-resolution images, but that’s expected with big datasets. Other than that, everything—from annotation tools to export options—has worked smoothly. Review collected by and hosted on G2.com.

Quick to implement on NVIDIA Jetson device, easy to train models, lot's of different training options, workflows with tons of tools to automate processes. Support team is quick to respond and help answer questions and resolve issues. Constantly innovating. Able to easily integrate with other tools we use. We use it every day for photo and video analysis and it is able to keep up as we scale. Review collected by and hosted on G2.com.

Wish it was easier to view and manage live stream video inference with their Growth subscription. Review collected by and hosted on G2.com.

Roboflow has become a key tool for computer vision research, especially in projects focused on real-time detection. What I like is its focus on simplifying the entire cycle for training models with datasets or your own datasets, which allows you to focus on data quality and algorithm optimization instead of wasting time on technical configurations. Review collected by and hosted on G2.com.

While Roboflow is a leading tool, the least useful feature is the integrated training tool for expert users. Auto-labeling requires a lot of human supervision (human-in-the-loop), and for very small objects or low-resolution images, Roboflow's AI often fails, requiring exhaustive manual labeling. Review collected by and hosted on G2.com.

This tool is easy to use, offers a wide variety of annotation styles, and includes an auto-annotation feature. It makes training easier, and the data augmentation feature is superb. Ease of integration with Google Colab. l also managed to get help when l wanted more raining credits. Student-oriented, can train as many times as you want, different models. Review collected by and hosted on G2.com.

If there were some datasets readily available, integrated into. Review collected by and hosted on G2.com.

As part of my graduation project, Camera AI System, an intelligent monitoring platform deployed across buildings, theme parks, and night markets, we needed to process massive amounts of video and image data from dozens of cameras. Our use cases required annotation of complex objects such as:

- Faces to differentiate employees, guests, VIPs, and unknown persons

- People under challenging conditions (masks, low light, backlight, crowd scenes)

- Vehicles (cars, buggies) entering restricted areas

- Special behaviors such as crowd gatherings, wrong-way movement, or intrusions Review collected by and hosted on G2.com.

There are some limitations with Research account Review collected by and hosted on G2.com.

Roboflow is a comprehensive platform that enables teams to collaborate on dataset management, data preparation, and model training in an integrated workflow. The platform is under active development, with new features and improvements released frequently. Review collected by and hosted on G2.com.

Roboflow’s API is not yet fully mature, which can limit automation and deeper integration into custom pipelines. In addition, feature support is uneven across project types: object detection workflows are significantly more developed than others, with keypoint detection in particular lacking the same level of functionality and tooling. Review collected by and hosted on G2.com.

What I like best about Roboflow is that it not only makes dataset preparation easy but also provides built-in model training and clear evaluation graphs that help me understand performance at a glance. I especially appreciate how simple it is to integrate the trained models into different applications, which makes the entire workflow—from data to deployment—smooth and enjoyable. Review collected by and hosted on G2.com.

What I dislike about Roboflow is that some of the best features and bigger dataset limits are only available in the paid plans, which can be tough for beginners or students who want to practice on larger projects. Review collected by and hosted on G2.com.