SuperAnnotate Reviews

I am a physician-scientist and my research applies computer vision techniques to biomedical imaging projects. The SuperAnnotate platform solves a common problem that computer vision researchers encounter, which is the rapid curation of expert annotations. Given the limited training data in medical imaging datasets, I often start with image segmentation based on expert annotations. Previously, I cobbled together several different tools for annotation, quality control, and neural network training but it was cumbersome to switch between different software and intermediate annotation masks. My favorite aspect of SuperAnnotate is the integrated, streamlined workflow AI-augmented annotation, quality control, and neural network training. Note: I have left a review on TrustPilot with similar content, although the wording has been changed to avoid self-plagiarism.

Updated review 2024-02-16:

I'm still using SuperAnnotate and it has been crucial to make progress on several projects. They have continued to innovate developing new annotation features such as MagicSelect and Magic polygon. The shortcuts for annotating images are intuitive and speed up my workflow. Their annotation teams have been timely and efficient, following detailed instructions. I have been very happy with their services.

My initial concern was that I could not upload annotation masks that were generated manually or on other platforms. However, the team rapidly introduced a feature to upload annotation masks for use in the platform, which meant I could leverage existing annotations from previous projects. The team is responsive to feedback and I am very pleased with my experience.

Updated review 2024-02-16:

The team has been responsive to feedback with continuous development and integration of new features. Uploading pre-annotations has been easy using their SDK. I don't have no dislikes or concerns.

I have been able to annotate image datasets much faster, which has made my research more productive. I like the project management interface that allows the delegation of annotation to specific team members as well as the ability to monitor annotation progress. The integrated neural network training is very helpful, too.

Offers a wide range of customizable features. Proactive and attentive customer support.

Objects and tags colors could also be customizable. When annotating an object with more the one selectio area, same name (polygon, box, etc) can be confusing.

Facilitates interaction with collaborators in different countries, allowing for time-saving in project management.

SuperAnnotate was extremely easy to use and has a lot of utilities to decrease the time annotating.

They make it easy to roll back or change mistakes you make while annotating.

Though I did not need much support, their customer service was super friendly and always responded quickly.

They also had exporting options that perfectly suited my needs i.e. in COCO Format.

The only issue I had was the accuracy of the auto-segmentation at times, but this was likely due to the low resolution of the images I was annotating.

SuperAnnotate is solving the crisis of good data. The accuracy of machine learning and AI is completely built around how good and how accurate the data being used is, this service is a major step in the right direction for proper and efficient creation of training data.

SuperAnnotate has a concise UI, which is easy to get the keypoint and friendly to beginners.

SuperAnnotate has some automatic labeling functions, saving a lot of time from the heavy labeling works.

SuperAnnotate provides a wonderful data management service. Different projects can be handled easily.

It will be better if the magic select function supports concating different part of selections to one object.

I have some images in autodriving to be labeled. These images will be used to train model for two tasks, detection and segmentation. It costs a lot of time to manually annotate these images. SuperAnnotate is a timesaver product with some autolabeling functions such as magic select.

1. The quality of the annotations we have received has been fantastic, with viritually no issues found.

2. SuperAnnotate has been very responsive with our questions and concerns on delivery timelines, which has been super helpful in meeting our deadlines.

3. They have a great team on the businesss and annotation side with straight answers to questions and overall, a great pleasure to work with!

There isn't anything we really dislike about SuperAnnotate.

Providing large scale annotated data to train ML models.

It's super easy to understand and use. It is practical and functional; It has all the right features I need to label my thin blood smear images. One of my major concern when choosing a labeling software was the privacy of my data and superannotate did great with that. The customer support experience is really great, their response is quick, efficient and effective.

There is nothing I dislike about SuperAnnotate

SuperAnnotate is aiding AI practitioners with image annotation, segmentation among others. My work is to develop deep learning models that would aid microscopists in diagnosing haematological diseases using images of microscopic thin blood smear. Thus, I am using SuperAnnotate to label;

1. Different species of malaria parasites and their life stages found in red blood cells

2. Different types of white blood cells

3. Different type of red blood cell abnormalities such as; rouleaux formation, bite cells, target cells among others

- The UI is simple and basic navigation is mostly intuitive.

- The customer support is very responsive.

- The product team is very open to suggestions and very interested in client feedback--they have actioned several requests originating from our use.

- Handles multi layer images very well.

- Built in analytics are useful and they continue to grow.

- Admin functionality is very limited. Managing users and projects is time consuming and requires many repetitive steps.

- Image annotation UX is geared towards more fewer, simple images (like street signs, etc.). Extensive and densley populated images are bit more tricky to annotate within their editor.

- Not enough customization or configuration options. Even through SDK, it is very limited on what you can change and modify.

- Documentation is often outdated and thus incorrect.

Ease of labeling training data and integrating that with an analytics database to store annotated data.