main.py 文件源码

def test(epoch):
    Net.eval()
    preditions=[]
    for iteration,(inputss,labelss) in enumerate(testing_data_loader,1):
        #embed()
        inputss=Variable(inputss.view(-1,3,training_size,training_size))
        if cuda:
            inputss=inputss.cuda()
        #embed()
        if model=='resnet101':
            prediction=Net.module.resnet101(inputss).cpu().data.numpy()
        elif model=='inception_v3':
            prediction=Net.module.inception_v3(inputss).cpu().data.numpy()
        elif model=='inception_v4':
            prediction=Net.module.inception_v4(inputss).cpu().data.numpy()
        prediction=prediction.mean(0).argmax()
        preditions.append(str(prediction))
        #print 'video num: ',iteration,' predition: ',str(prediction)
    with open('/S2/MI/zqj/video_classification/data/ucf101/tmp_result/{}result_{}_new_epoch'.format(save_prefix,model)+str(epoch)+'.txt','w')as fw:
        fw.write('\n'.join(preditions))
    str_out='python compute_test_result.py {}result_{}_new_epoch'.format(save_prefix,model)+str(epoch)+'.txt'
    os.system(str_out)