# ESC-50音频事件识别项目
基于tensorflow 2.7.0进行音频事件识别, 使用ESC-50数据集
## 模型介绍
包括自己搭建的CNN模型以及CRNN模型
## tflite模型
利用tenflow-lite保存为移动端的模型结构，对模型进行了压缩，包括不同的量化与剪枝方法
## 模型评估
对压缩前后不同模型在测试集上的效果进行了评估
## 今后目标
将tflite模型部署到手机上测试压缩前后的性能收益
基于Tensorflow模型使用CNN与CRNN模型对ESC50数据集进行音频事件检测，结果导出为tflite，并进行剪枝量化 （180个子文件） 
metrics.csv 5KB metrics.csv 3KB .gitattributes 320B .gitignore 182B esc50_.18_0.6285_0.8050.hdf5 3.92MB esc50_.09_0.7725_0.7900.hdf5 3.92MB esc50_.11_0.7010_0.7775.hdf5 3.92MB esc50_.38_0.5391_0.8375.hdf5 3.92MB esc50_.03_1.3639_0.5875.hdf5 3.92MB esc50_.02_1.6943_0.5275.hdf5 3.92MB esc50_.01_2.7630_0.2425.hdf5 3.92MB esc50_.36_0.5500_0.8350.hdf5 3.92MB esc50_.17_0.6198_0.8375.hdf5 3.92MB esc50_.21_0.5927_0.8200.hdf5 3.92MB esc50_.31_0.5717_0.8325.hdf5 3.92MB esc50_.32_0.5289_0.8450.hdf5 3.92MB best.hdf5 3.92MB esc50_.29_0.5459_0.8500.hdf5 3.92MB esc50_.06_1.0149_0.7150.hdf5 3.92MB esc50_.04_1.1429_0.6575.hdf5 3.92MB esc50_.24_0.5608_0.8300.hdf5 3.92MB esc50_.30_0.5470_0.8450.hdf5 3.92MB esc50_.13_0.7288_0.7975.hdf5 3.92MB esc50_.27_0.5624_0.8325.hdf5 3.92MB esc50_.35_0.5350_0.8450.hdf5 3.92MB esc50_.20_0.5785_0.8325.hdf5 3.92MB esc50_.14_0.7228_0.7900.hdf5 3.92MB esc50_.28_0.5281_0.8525.hdf5 3.92MB esc50_.19_0.5972_0.8325.hdf5 3.92MB esc50_.23_0.5733_0.8475.hdf5 3.92MB esc50_.12_0.7720_0.7725.hdf5 3.92MB esc50_.15_0.6017_0.8425.hdf5 3.92MB esc50_.08_0.7809_0.7875.hdf5 3.92MB esc50_.34_0.5438_0.8475.hdf5 3.92MB esc50_.37_0.5394_0.8450.hdf5 3.92MB esc50_.10_0.8041_0.7575.hdf5 3.92MB esc50_.16_0.6201_0.8225.hdf5 3.92MB esc50_.26_0.5579_0.8375.hdf5 3.92MB esc50_.25_0.5622_0.8275.hdf5 3.92MB esc50_.05_0.9562_0.7075.hdf5 3.92MB esc50_.22_0.5599_0.8325.hdf5 3.92MB esc50_.33_0.5387_0.8350.hdf5 3.92MB esc50_.07_0.8790_0.7400.hdf5 3.92MB esc50_.01_3.1646_0.1775.hdf5 2.31MB esc50_.02_2.5710_0.2300.hdf5 2.31MB esc50_.52_1.0576_0.6575.hdf5 2.31MB esc50_.43_1.0932_0.6775.hdf5 2.31MB esc50_.29_1.1700_0.6450.hdf5 2.31MB esc50_.34_1.2442_0.6350.hdf5 2.31MB esc50_.40_1.1421_0.6700.hdf5 2.31MB esc50_.49_1.0457_0.6800.hdf5 2.31MB esc50_.27_1.2678_0.6225.hdf5 2.31MB esc50_.18_1.4221_0.5875.hdf5 2.31MB esc50_.30_1.3373_0.5950.hdf5 2.31MB esc50_.55_1.0632_0.6575.hdf5 2.31MB esc50_.47_1.0894_0.6600.hdf5 2.31MB esc50_.39_1.1113_0.6575.hdf5 2.31MB esc50_.03_2.3773_0.2850.hdf5 2.31MB esc50_.60_1.0327_0.6800.hdf5 2.31MB esc50_.16_1.4412_0.5650.hdf5 2.31MB esc50_.58_1.0367_0.6675.hdf5 2.31MB esc50_.64_1.0534_0.6575.hdf5 2.31MB esc50_.10_1.5370_0.5575.hdf5 2.31MB esc50_.59_1.0364_0.6700.hdf5 2.31MB esc50_.51_1.0523_0.6525.hdf5 2.31MB esc50_.07_1.8469_0.4550.hdf5 2.31MB esc50_.42_1.1583_0.6425.hdf5 2.31MB esc50_.31_1.2024_0.6325.hdf5 2.31MB esc50_.06_2.0109_0.3800.hdf5 2.31MB best.hdf5 2.31MB esc50_.61_1.0430_0.6600.hdf5 2.31MB esc50_.63_1.0405_0.6625.hdf5 2.31MB esc50_.32_1.1584_0.6475.hdf5 2.31MB esc50_.62_1.0292_0.6725.hdf5 2.31MB esc50_.41_1.1138_0.6875.hdf5 2.31MB esc50_.53_1.0327_0.6725.hdf5 2.31MB esc50_.23_1.2667_0.6075.hdf5 2.31MB esc50_.56_1.0450_0.6675.hdf5 2.31MB esc50_.21_1.2848_0.6075.hdf5 2.31MB esc50_.14_1.5356_0.5725.hdf5 2.31MB esc50_.15_1.4910_0.5700.hdf5 2.31MB esc50_.09_1.6796_0.5075.hdf5 2.31MB esc50_.05_1.9589_0.4550.hdf5 2.31MB esc50_.17_1.3278_0.6100.hdf5 2.31MB esc50_.26_1.2142_0.6475.hdf5 2.31MB esc50_.08_1.7489_0.4725.hdf5 2.31MB esc50_.35_1.1851_0.6525.hdf5 2.31MB esc50_.45_1.0374_0.7000.hdf5 2.31MB esc50_.50_1.0702_0.6700.hdf5 2.31MB esc50_.22_1.3197_0.6125.hdf5 2.31MB esc50_.24_1.2920_0.6175.hdf5 2.31MB esc50_.28_1.2442_0.6225.hdf5 2.31MB esc50_.57_1.0226_0.6850.hdf5 2.31MB esc50_.44_1.0495_0.6925.hdf5 2.31MB esc50_.36_1.1031_0.6700.hdf5 2.31MB esc50_.33_1.1566_0.6475.hdf5 2.31MB esc50_.13_1.4391_0.5800.hdf5 2.31MB esc50_.54_1.0155_0.6800.hdf5 2.31MB esc50_.04_2.0656_0.3975.hdf5 2.31MB esc50_.38_1.2612_0.6350.hdf5 2.31MB共 180 条
- 1
- 2
资源评论
