mobilenet-v1-ssd300.tensorflow (8bit quantized and fine-tuned)

Application: Single-stage Object Detection Base model: MobileNet-v1 Framework: tensorflow1.1 Training Information: weights were fine-tuned using TF fake quantization nodes Quality: The COCO mAP(IoU=0.50:0.95) on 5000 validation images is 23.0% Precision: 8-bit precision Is Quantized: Yes, using fake quantization - i.e., weights appear in float32 but have only 256 unique values. Dataset: COCO val-2017