# Also Detecting the Not Trained Object

**URL:** <https://discuss.roboflow.com/t/also-detecting-the-not-trained-object/4294>\
**Category:** 🤝  Community Help\
**Created:** [January 25, 2024, 8:23am UTC](https://discuss.roboflow.com/t/also-detecting-the-not-trained-object/4294 "2024-01-25T08:23:27Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![Kaung\_Khant\_Zaw](https://yyz1.discourse-cdn.com/flex029/user_avatar/discuss.roboflow.com/kaung_khant_zaw/32/3674_2.png) [@Kaung\_Khant\_Zaw](https://discuss.roboflow.com/u/Kaung_Khant_Zaw)\
**Post date:** [January 25, 2024, 8:23am UTC](https://discuss.roboflow.com/t/also-detecting-the-not-trained-object/4294/1 "2024-01-25T08:23:27Z")

</div>

![Screenshot 2024-01-25 144910](https://canada1.discourse-cdn.com/flex029/uploads/roboflow1/original/2X/8/82395ffd7b6827b489e3c6b594dc3721284ea517.jpeg)

In the image, I only trained the brand text but selected the whole caps. After trained, I test it in the roboflow live webcam and it also detects the flip cap. How can I solve it? Should I train only the design inside the cap?

---

<div class="post-metadata">

**Author:** ![leo](https://yyz1.discourse-cdn.com/flex029/user_avatar/discuss.roboflow.com/leo/32/2228_2.png) [@leo](https://discuss.roboflow.com/u/leo)\
**Post date:** [January 31, 2024, 4:29am UTC](https://discuss.roboflow.com/t/also-detecting-the-not-trained-object/4294/2 "2024-01-31T04:29:29Z")

</div>

Hi @Kaung_Khant_Zaw

Computer vision models are trained to detect things that are similar to the ones they are trained with. Your model needs to learn that there’s a distinction to be made between the two. Usually, more training data solves this problem:

- Add more images that include your “flipped cap”, then:
  - Make sure not to annotate them, making then [a null annotation](https://blog.roboflow.com/missing-and-null-image-annotations/#:~:text=A%20null%20annotation%20occurs%20when%20an%20image%20has%20no%20objects%20present%2C%20and%20thus%2C%20no%20bounding%20boxes%20need%20to%20be%20recorded.%20This%20is%20not%20necessarily%20problematic%20%E2%80%93%20in%20fact%2C%20it%20may%20be%20desired%20in%20order%20to%20train%20a%20model%20that%20objects%20are%20not%20always%20present%20in%20frame)
  - Or label them as a different class from the object that you want to detect.

---

<div class="post-metadata">

**Author:** ![system](https://canada1.discourse-cdn.com/flex029/uploads/roboflow1/original/1X/bb6edbfa56d0cf2ee603f061e9e09103502cc8cc.png) [@system](https://discuss.roboflow.com/u/system)\
**Post date:** [February 21, 2024, 4:29am UTC](https://discuss.roboflow.com/t/also-detecting-the-not-trained-object/4294/3 "2024-02-21T04:29:36Z")

</div>

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.