CorricLabs

Part Three

Q1. Collect Dataset/Photos

I collected some images from different locations and scenarios to really test how well the algorithm worked. I have a few extra photos from a holiday in Thailand that I wanted to include to see how well the algorithm worked in busy environments.


Q2. Test the Prediction Performance

Each image after being processed by the YOLO object detection system using the images as the inputs.


Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Image 21
Image 22
Image 23
Image 24
Image 25
Image 26
Image 27
Image 28
Image 29
Image 30
Image 31
Image 32
Image 33
Extra Image 1
Extra Image 2

Q3. Evaluation

Analyse the prediction results and write an evaluation report. Summarise the prediction accuracies by filling up the table.

Transportation type Total number appeared in all images Total number accurately detected in all images Average accuracy (%)
Cars 163 154 94.48%
Bikes 8 6 75%
Trucks 11 11 100%
People 11 11 100%

The algorithm worked much better than I originally expected, correctly detecting 94.3% of relevant objects in the images.

The algorithm correctly detected 154/163 cars (94.48%), 6/8 bikes (75%), 11/11 trucks (100%) and 11/11 people (100%).

Occasionally the algorithm would incorrectly detect cars or trailers as trucks, however this was rare. The algorithm did have some trouble making out and detecting a few of the motorbikes in the images from Phuket, I imagine due to the amount of content in the images. However, the algorithm was still impressive in these scenarios, correctly detecting 92% of relevant objects in the images.