Base Model

[ry-eon]

Base Model

Model

• base network : Unet
• encoder : mobilenet_V2
• pretrained : imagenet
• image : issue

Comparing resnet34, efficientnet-b0, mobilenet_v2, I find that 4-10 epochs are sufficiently trained. Therefore, I want to reduce the inference time by selecting moblienet_v2 with the fewest parameters.

Augmentation

1. Total Process

2. Code Implementation

• Copy to Simple Random Location
• Copy to Transparent Random Location
• Copy to Color Random Location
• Copy to Color + Transparent Random Location

3. Performance Evaluation

• Copy to Simple Random Location
• Copy to Transparent Random Location
• Copy to Color Random Location
• Copy to Color + Transparent Random Location