Microsoft

Custom Vision Service

Welcome to the Custom Vision Service API Forum

Categories

API – Any ideas or feedback pertaining to features or enhancements to Custom Vision Service API.

Documentation – Any ideas or suggestions for the API Reference or Documentation.

Language Support – Submit a request to have a particular language supported.

Samples & SDK Request – Let us know if you would like to see a Code sample or SDK provided.

Release Notes – Recently updated December 19.

How can we improve Custom Vision Service?

(thinking…)

Enter your idea and we'll search to see if someone has already suggested it.

If a similar idea already exists, you can support and comment on it.

If it doesn't exist, you can post your idea so others can support it.

Enter your idea and we'll search to see if someone has already suggested it.

  1. Increase Max image size (prediction)

    Most cameras and smart phones will produce pictures bigger than the Max image size (prediction) in the Vision Service API. Could it be increased or create a simple way to send a rescaled image to the Vision service API.

    1 vote
    Sign in
    Check!
    (thinking…)
    Reset
    or sign in with
    • facebook
    • google
      Password icon
      Signed in as (Sign out)

      We’ll send you updates on this idea

      0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
    • support 1000 tags

      I would like to add about 1000 tags, is this possible?

      1 vote
      Sign in
      Check!
      (thinking…)
      Reset
      or sign in with
      • facebook
      • google
        Password icon
        Signed in as (Sign out)

        We’ll send you updates on this idea

        0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
      • The difference between model in the cloud and on-prem

        Hello PM,

        I'm working on an urgent case using custom vision to do the selection of defected solar cell photos from a huge amount of library. Due to security requirement, the customer wants to export the model to on-prem to do the prediction. According to the documentation, it could be slightly less accurate than doing in the cloud. I'm wondering if you have the further explanation regarding this difference and how to improve the precision on-prem. Another question is how will it be charged on-prem prediction if the server can't connect to Azure. Hope to hear the feedback soon.

        1 vote
        Sign in
        Check!
        (thinking…)
        Reset
        or sign in with
        • facebook
        • google
          Password icon
          Signed in as (Sign out)

          We’ll send you updates on this idea

          0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
        • Export Object Detection

          It would be great if one could export models to iOS and other platforms while using a different domain than just General (compact) or even Object Detection.

          12 votes
          Sign in
          Check!
          (thinking…)
          Reset
          or sign in with
          • facebook
          • google
            Password icon
            Signed in as (Sign out)

            We’ll send you updates on this idea

            2 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
          • C# Sample Code should check if a project exists instead of creating another project of the same name

            As I tried the sample app, after running it twice, the thrid time it resulted in an error. Turned out it has exceed the max project limit. The confusion was that it created two. I assumed that it just overwrote the project. A minor issue, but might be confusing that the second time it was another project, and third time it failed without much clear explanation why.

            1 vote
            Sign in
            Check!
            (thinking…)
            Reset
            or sign in with
            • facebook
            • google
              Password icon
              Signed in as (Sign out)

              We’ll send you updates on this idea

              0 comments  ·  Samples & SDKs  ·  Flag idea as inappropriate…  ·  Admin →
            • Improve Account Selection

              On my https://customvision.ai/projects, it is now showing all my Azure accounts. (I swear it did not this weekend). But it is a bit confusing to see the "email" address labelled with onmicrosoft.com address.

              2 votes
              Sign in
              Check!
              (thinking…)
              Reset
              or sign in with
              • facebook
              • google
                Password icon
                Signed in as (Sign out)

                We’ll send you updates on this idea

                0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
              • tag

                if i map one product for 1 tag (SKY number), i wish to predict which products are most similar to the user submitted image, there is a limit of 250 tag/project.
                it seems i can't handle more than 1000 products, right?

                2 votes
                Sign in
                Check!
                (thinking…)
                Reset
                or sign in with
                • facebook
                • google
                  Password icon
                  Signed in as (Sign out)

                  We’ll send you updates on this idea

                  1 comment  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                • Select All on Predictions view

                  I would like to be able to Select All images on the current page so that I can then use the "Tag Images" button.

                  Currently, I have to click on each individual checkbox to select them all.

                  2 votes
                  Sign in
                  Check!
                  (thinking…)
                  Reset
                  or sign in with
                  • facebook
                  • google
                    Password icon
                    Signed in as (Sign out)

                    We’ll send you updates on this idea

                    0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                  • Thumbnails are cropped, so I can't bulk-tag (I can't tag without seeing all 4 corners of image)

                    On the image list pages ("Training Images" and "Predictions" tabs), the thumbnails are cropped to be square ratio. But this means I can't see the whole image.

                    The "Tag Images" button is very useful, allowing me to tag images from the list view. But as a human doing the tagging, I need to be able to see the whole image.

                    Possible solutions:
                    1. Use zooming instead of cropping when generating the thumbnail. Just pad with white or black to make the square.

                    2. Add a toggle so the user can set whether they want zoomed or cropped thumbnails.

                    1 vote
                    Sign in
                    Check!
                    (thinking…)
                    Reset
                    or sign in with
                    • facebook
                    • google
                      Password icon
                      Signed in as (Sign out)

                      We’ll send you updates on this idea

                      0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                    • Mobile app to snap and tag photos

                      Anno is a mobile and web app that makes the process of adding and tagging photos in CustomVision projects super easy. The app isn't perfect, but it's free to use and works quite well with CustomVision. Looks like additional functionality is coming soon. annoapp.com

                      1 vote
                      Sign in
                      Check!
                      (thinking…)
                      Reset
                      or sign in with
                      • facebook
                      • google
                        Password icon
                        Signed in as (Sign out)

                        We’ll send you updates on this idea

                        0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                      • Give more information about prefered image size

                        I have experimented with original and resized images to train classifiers and got strange results.

                        When I used the "General" domain, the performance scores were better for original size images (5Mpixel).
                        When I trained "General (Compact)" the performance score were better for resized (1Mpixel - width/height: 768*1024px) images.

                        Can you give more infomation about what size do you use as input to train your models? Can the image resize algorythm cause the difference between performance of the same model?

                        1 vote
                        Sign in
                        Check!
                        (thinking…)
                        Reset
                        or sign in with
                        • facebook
                        • google
                          Password icon
                          Signed in as (Sign out)

                          We’ll send you updates on this idea

                          0 comments  ·  Documentation  ·  Flag idea as inappropriate…  ·  Admin →
                        • Allow Project Thumbnail Editing.

                          It would be helpful if we could select a specific image as a representative thumbnail of the project. The current random image is confusing when viewing all the projects on the main page.

                          2 votes
                          Sign in
                          Check!
                          (thinking…)
                          Reset
                          or sign in with
                          • facebook
                          • google
                            Password icon
                            Signed in as (Sign out)

                            We’ll send you updates on this idea

                            0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                          • 1 vote
                            Sign in
                            Check!
                            (thinking…)
                            Reset
                            or sign in with
                            • facebook
                            • google
                              Password icon
                              Signed in as (Sign out)

                              We’ll send you updates on this idea

                              0 comments  ·  Documentation  ·  Flag idea as inappropriate…  ·  Admin →
                            • Train with Random cropping - Predict with multiple cropping and majority voting

                              Have a feature to switch on cropping (data augmentation) for training and for inference.

                              1) It is a well known technique to use random cropping to help to train a classifier. (Alex Krizhevsky did that too)

                              2) In this paper https://arxiv.org/pdf/1803.00949v1.pdf about Tree Species Identification from Bark Images Using Convolutional
                              Neural Networks researchers used cropping for training and for interference too. " For
                              the latter, we split the test image into multiple non-overlapping
                              224x224 crops and classified each one individually. Then, we
                              performed majority voting to determine the final outcome.
                              As can be seen from Table III, progressing from single …

                              2 votes
                              Sign in
                              Check!
                              (thinking…)
                              Reset
                              or sign in with
                              • facebook
                              • google
                                Password icon
                                Signed in as (Sign out)

                                We’ll send you updates on this idea

                                0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                              • need documentation on using testing and training key API in code

                                Please provide some documentation on using the testing and training key API in code. Also do we need a end url? Documentation to use this in code is sorely missing.

                                1 vote
                                Sign in
                                Check!
                                (thinking…)
                                Reset
                                or sign in with
                                • facebook
                                • google
                                  Password icon
                                  Signed in as (Sign out)

                                  We’ll send you updates on this idea

                                  0 comments  ·  Documentation  ·  Flag idea as inappropriate…  ·  Admin →
                                • 2 votes
                                  Sign in
                                  Check!
                                  (thinking…)
                                  Reset
                                  or sign in with
                                  • facebook
                                  • google
                                    Password icon
                                    Signed in as (Sign out)

                                    We’ll send you updates on this idea

                                    1 comment  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                                  • Download the TensorFlow or CoreML using the API

                                    Possibility to download the TensorFlow or CoreML model from the API. I don't want to publish a new App every time I need to update the model. There is any recommendation way to update the App model without new publish?
                                    Thanks

                                    2 votes
                                    Sign in
                                    Check!
                                    (thinking…)
                                    Reset
                                    or sign in with
                                    • facebook
                                    • google
                                      Password icon
                                      Signed in as (Sign out)

                                      We’ll send you updates on this idea

                                      0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                                    • i don't want the data to be public

                                      my data is sensitive so don't want to upload images to custom vision site.i have a azure subscription so can i use it to make as private???

                                      1 vote
                                      Sign in
                                      Check!
                                      (thinking…)
                                      Reset
                                      or sign in with
                                      • facebook
                                      • google
                                        Password icon
                                        Signed in as (Sign out)

                                        We’ll send you updates on this idea

                                        0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                                      • Download Project / Images

                                        Please add a download as zip in web interface from customvision.ai for

                                        - export project
                                        - import project
                                        - training images / iteration
                                        - predicted images / iteration
                                        - red boxed(the ones that are not fully recognized) images / iteration

                                        Thanks in advance

                                        2 votes
                                        Sign in
                                        Check!
                                        (thinking…)
                                        Reset
                                        or sign in with
                                        • facebook
                                        • google
                                          Password icon
                                          Signed in as (Sign out)

                                          We’ll send you updates on this idea

                                          0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                                        • one hot encoding

                                          There needs to be an option for one-hot encoding so the output is the probability of one answer over the other.

                                          1 vote
                                          Sign in
                                          Check!
                                          (thinking…)
                                          Reset
                                          or sign in with
                                          • facebook
                                          • google
                                            Password icon
                                            Signed in as (Sign out)

                                            We’ll send you updates on this idea

                                            0 comments  ·  API  ·  Flag idea as inappropriate…  ·  Admin →
                                          ← Previous 1 3
                                          • Don't see your idea?

                                          Custom Vision Service

                                          Feedback and Knowledge Base