add results

Author: HiKapok

README.md

@@ -4,7 +4,7 @@ This repository contains codes of the reimplementation of [SSD: Single Shot Mult
 
 There are already some TensorFlow based SSD reimplementation codes on GitHub, the main special features of this repo inlcudes:
 
-- state of the art performance when training from VGG-16 pre-trained model (trained model will be released later).
+- state of the art performance(~77%mAP) when training from VGG-16 pre-trained model (SSD300-VGG16).
 - the model is trained using TensorFlow high level API [tf.estimator](https://www.tensorflow.org/api_docs/python/tf/estimator/Estimator). Although TensorFlow provides many APIs, the Estimator API is highly recommended to yield scalable, high-performance models. 
 - all codes were writen by TensorFlow ops (no numpy operation) to ensure the performance and portability.
 - using ssd augmentation pipeline discribed in the original paper.
@@ -55,5 +55,26 @@ All the codes was tested under TensorFlow 1.6, Python 3.5, Ubuntu 16.04 with CUD
 
 This repo is just created recently, any contribution will be welcomed.
 
+## Results (VOC07 Metric)
+
+This implementation(SSD300-VGG16) yield **mAP 76.94%** on PASCAL VOC 2007 test dataset, the details are as follows:
+
+| sofa   | bird  | pottedplant | bus | diningtable | cow | bottle | horse | aeroplane | motorbike
+|:-------|:-----:|:-------:|:-------:|:-------:|:-------:|:-------:|:-------:|:-------:|:-------:|
+|  80.6  |  75.3 |  51.8   |   85.1  |   77.1    |  81.7 |  49.3  | 85.5  |   80.1    |   83.9   |
+| sheep  | train | boat    | bicycle | chair    | cat   | tvmonitor | person | car  | dog |
+|  79.1  |  86.4 |  70.3   |   82.5  |   61.9    | 87.8 |  73.7  | 78.5  |   82.7   |   85.5   |
+
+You can download the trained model(VOC07+12 Train) from [GoogleDrive](https://drive.google.com/open?id=1yeYcfcOURcZ4DaElEn9C2xY1NymGzG5W) for further research.
+
+Here is the training logs and some detection results:
+
+![](logs/loss.jpg "loss")
+![](logs/celoss.jpg "celoss")
+![](logs/locloss.jpg "locloss")
+![](demo/demo1.jpg "demo1")
+![](demo/demo2.jpg "demo2")
+![](demo/demo3.jpg "demo3")
+
 ## ##
 Apache License, Version 2.0