Iterate a Model
If the model performs poorly after deployment, it can be improved through iteration. The traditional approach is to retrain the model by adding more data. Alternatively, you can use model finetuning to save training time while maintaining the current accuracy, enabling more efficient model iteration.
| This feature only works under the Developer Mode. |
Normal Model Iteration
Models you trained can be fine-tuned using the following steps:
-
Acquire images that report poor recognition results.
-
Use Mech-DLK to open the project the model belongs to.
-
Enable the Developer mode by clicking .
-
Import the acquired images into the corresponding module of the model and complete labeling.
-
On the Training tab, click the expand icon
to the right of Training Parameter Settings. In the Training Parameter Settings window, under the Model finetuning tab, enable Finetune. (For the Instance Segmentation module, you also need to select Self finetuning.) -
Switch to the Training parameters tab, and reduce Epochs to 50–80.
-
Click OK to save the parameter settings.
|
Fast Positioning, Pick Anything V2, and Object-Bin Segmentation modules do not support model finetuning settings in the Training Parameter Settings window. To improve performance of these models, you can add more data and retrain the model. |
General Model Iteration
General models are models developed by Mech-Mind. For more information about the general models and instructions on using general models, see General Models.
You can use the following method to fine-tune the model:
-
Acquire images that report poor recognition results.
-
Open Mech-DLK, create a new project, and add the corresponding module.
-
Enable the Developer mode by clicking .
-
Import the acquired images into the corresponding module of the model and complete labeling.
-
On the Training tab, click the expand icon
to the right of Training Parameter Settings. In the Training Parameter Settings window, under the Model finetuning tab, enable Finetune and select Deep learning model finetuning, then click the folder icon 
to choose a general model (a .dlkmp file).The Deep learning model finetuning feature can be used to fine-tune the general model.
-
Switch to the Training parameters tab, and reduce Epochs to 50–80.