[Technical Information] Is it possible to distinguish collected images other than the operating environment as training data?

Added to bookmarks
Bookmarks list

Bookmark has been removed
Bookmarks list

You can't add any more bookmarks

By registering as a member, you can increase the number of bookmarks you can save and organize them with labels.
Free membership registration

DOBOT Magician

last updated：Jan 09, 2025

Official site

Sample program available ■ We verified whether it is possible to distinguish images that were not taken in a real operating environment using a smartphone camera as training data!

We would like to introduce a technical verification article on the tabletop-sized small robotic arm, DOBOT Magician. ■ Article here https://qiita.com/shimamotosan/items/1a0c6c73b1e60ae5db43 ■ Content We are collecting learning data for objects to be identified in an environment where sorting is performed. However, in AI-based image recognition, it is possible to include image data collected outside the actual operational environment in the training. This time, we verified whether it is possible to identify objects using images taken with a smartphone camera that are not from the actual operational environment as training data. Keywords: Python, OpenCV, AI, image recognition, dobot ■ About DOBOT Magician DOBOT Magician is a robotic arm equipped with a 4-axis arm, priced around 150,000 yen. The main unit does not require prior safety training, making it less restricted in terms of usage locations and applicable in various scenes. * DOBOT Magician product introduction page https://afrel.co.jp/product/dobot

Related Link - https://afrel.co.jp/product/dobot

Inquire About This Product

basic information

■DOBOT Magician * Main unit: Weight 3.4kg * 4-axis arm * Position repeatability: 0.2mm * Dedicated software (DobotStudio): Free to use * Tool head (included as standard) - Suction cup kit - Pneumatic gripper (can lift objects up to approximately 500g) - Lighting & drawing kit (includes a dedicated pen) - 3D printing kit * Functions - Gripping - Teach & playback (records multiple XYZ coordinates to form a sequence. Can reproduce movements.) * Features - Graphical programming (programming by simply arranging blocks) - Expandable I/O (functionality can be expanded by connecting external devices such as sensors using 13 expansion ports) - Connect a web camera for image recognition/processing and robot arm control * Optional products - Conveyor belt Includes photoelectric switches, color sensors, and wooden blocks, allowing for learning and experimentation that replicates a production line - Slider rail kit Uses slider rails over 1m to enable work while moving horizontally

Price information

DOBOT Magician Educational Version: 226,000 yen excluding tax DOBOT Magician Standard Version: 180,000 yen excluding tax

Price range

Delivery Time

Please contact us for details

※Delivery times may vary depending on the quantity, so please contact us for more information.

Applications/Examples of results

- Examination and demonstration aimed at streamlining and automating production lines in the manufacturing industry Movement of parts (workpieces) Sorting of parts using image recognition and processing - Control learning of robotic arms - Utilization in production management and product development departments * A representative introduction pattern is presented in the explanatory article. https://learninglab.afrel.co.jp/1145/

Line up（4）

Model number	overview
AFD-ED21: DOBOT Magician Educational Version	The Standard Version comes with peripheral parts such as a Bluetooth module, Wi-Fi module, and joystick control kit.
AFD-ST38: DOBOT Magician Standard Version	The robot arm itself comes with a suction cup kit, gripper, writing and drawing kit, and a 3D printing kit.
AFD-CV8: Conveyor Belt	Includes the main unit, photoelectric switch, color sensor, and a set of wooden blocks.
AFD-SLRK: Slider Rail Kit	Includes a slider rail main unit with a length of over 1m, a wire set, and a toolkit.