There are two things that you can evaluate for any methods, not only for segmentation: accuracy and time execution. For image segmentation, for accuracy, there are a lot of metrics. For the performance, I think you are focusing on the time execution. So, you just evaluate the total time of implementation of your method.

I need for both but mostly like want acuracy@dang

If for accuracy, I think you should evaluate dice score, jaccard score, F1 score and accuracy (precision) and recall. All values can be handled by matlab with function dice, jaccard, bfscore.

Segmentation is calculated using the DICE, which is defined by:

DICE=2∣∣S1g∩S1t∣∣∣∣S1g∣∣+∣∣S1t∣∣=2TP2TP+FP+FN

OR the Jaccard index (JAC) between two sets is defined as the intersection between them divided by their union, that is:

JAC=∣∣S1g∩S1t∣∣∣∣S1g∪S1t∣∣=TPTP+FP+FN

The performance evaluation should be quantitative. It should report how many objects were detected correctly and how many false positives (false alarms) were produced. It should support one-to-one matches, one to many matches, and many to one matches, and the evaluation should scale up to larger test areas or multiple 3D scenes without losing its tracking capability.

Pullarao Chennamsetty

Dice and Jaccard metrics are the most well-know measures for segmentation accuracy evaluation. However, you must have the ground truth / gold standard segmentation. Yet, there is another way to evaluate segmentation accuracy if there is NO ground truth via using some entropy metrics of the segmented regions.

Best

DICE index, Jaccard Index. You can read following paper for more details about segmentation evaluation metrics:

Article Automatic Deep Feature Learning via Patch-Based Deep Belief ...

There are several, the comparision of the segemntation result with the produced by a human expert is the best

Indeed, depending on the code performance if good code did not have any error, so please depend on your code and which program use fort hat.