Segments AI

What do you like best about Segments.ai?

I am a master's student, and I couldn't find a good tool for 3D point cloud annotation until I found Segments.ai. They generously provided me with a free academic license. This tool was exactly what I needed. It is easy to use, accepts general formats like PCD (not just BIN formats), and they send helpful emails and Google Colab notebooks.

What do you dislike about Segments.ai?

There isn't much to dislike about Segments.ai. I wish they made more videos on YouTube as they are very helpful.

What problems is Segments.ai solving and how is that benefiting you?

Segments.ai solves the problem of 3D point cloud annotation, which is essential for my research and projects.

What do you like best about Segments.ai?

The best thing about segments.ai is how fast one can label images. With other tools that I have used it could take up to 5 minutes to label one image. But with segments.ai it can take as little as 10-20 seconds to label one image with several classes.

The reason for why it is so easy to label images with segments.ai is that this product comes with several different features such as edit with superpixels, brush, polygon etc and it also comes with autosegmentation (which as of now is in beta mode).

Moreover, as a manager it is also very easy to se statistics for each labeler and this product also has a very nice review function so that you easily can correct mistakes. However, mistake seldomly happens, since this tool make it so easy to label iamges.

Furthermore, once you are done with a labeling project there is an amazing python SDK that makes it very easy to download both images and segmentation bitmaps.

I would also like to add that when I encounter a problems with segments.ai (which rarely happens), their team are very quick to respond and they are extremely helpful.

What do you dislike about Segments.ai?

So far I have not encountered anything that I dislike about segments.ai

What problems is Segments.ai solving and how is that benefiting you?

As I mentioned before data labeling is a time consuming process, and with other tools that I have used before it couuld take me up to 5 minutes to label one image. However, with segments.ai it can take as little as 10-20 seconds to label one image with several classes.

I would also like to add that once you are done with a labeling project there is an amazing python SDK that makes it very easy to download both images and segmentatiojn bitmaps.

What do you like best about Segments.ai?

An easy interface and export function. Free academic access.

What do you dislike about Segments.ai?

It would be nice to have options to export annotations straight from the web version, without using Pythin SDA.

What problems is Segments.ai solving and how is that benefiting you?

Fastness of data annotation, accelerating my testing cycle for my research work.

What do you like best about Segments.ai?

I love all its features and how easy it is for both 2D and 3D labelling. The dataset creation and export is simple and intuitive. Labelling, especially in 2D, relies on many tools (superpixels, brushes, etc) that make the experience easy and straightforward to explain to an inexperienced person.

What do you dislike about Segments.ai?

The 3D visualization gets stuck at times because it's very computationally heavy. It would be cool to introduce a feature for "hierarchical labeling", meaning to have multiple instance ids for the same pixel/point. This is useful, for example, in the agricultural domain where both plant instance and leaf instance segmentation are useful, thus each pixel has to be associated to a specific plant and to a specific leaf.

What problems is Segments.ai solving and how is that benefiting you?

I mostly use Segments.ai to label my own datasets for segmentation tasks, and to supervise people labelling for me and reviewing their labels. It's tremendously useful, especially when scenes are not too cluttered (in which case the superpixel segmentation gets a bit messy, and probably a polygon strategy would be better-suited).

What do you like best about Segments.ai?

Versatile Annotation Options: The diverse range of annotation types, from bounding boxes to keypoints, empowers me to accurately label various types of data, ensuring comprehensive annotation for my projects.

AI-Assisted Labeling: The AI-powered pre-labeling feature is a true time-saver. It not only speeds up the annotation process but also enhances accuracy by intelligently predicting labels, elevating the quality of my annotated datasets.

What do you dislike about Segments.ai?

Complexity for New Users: Data annotation tools can sometimes have a learning curve, especially for users who are new to the concept of data labeling or have limited technical expertise. Navigating the platform, understanding annotation types, and using advanced features might be challenging initially.

Quality Control: Ensuring the accuracy and consistency of annotations can be a challenge, especially when dealing with complex or large datasets. Users might face difficulties in maintaining high-quality annotations across different annotators or addressing discrepancies in labeling.

What problems is Segments.ai solving and how is that benefiting you?

Time-Intensive Data Annotation: Manually labeling and annotating datasets can be time-consuming and resource-intensive.

Annotation Consistency and Accuracy: Maintaining consistent and accurate annotations across different annotators can be challenging and lead to discrepancies in labeled data.