pruneVGG.py 文件源码

def eval_accuracy_loss(X_data, y_data, BATCH_SIZE, top1_accuracy,top5_accuracy,loss_operation,images,y,RC,train_mode,regConst):
    nImgs = len(X_data)
    total_top1 = 0.0
    total_top5 = 0.0
    total_crossEn = 0.0
    sess = tf.get_default_session()
    for offset in range(0, nImgs, BATCH_SIZE):
        batch_x = utils.load_image(X_data[offset:offset+BATCH_SIZE])
        batch_y = y_data[offset:offset+BATCH_SIZE]
        t1,t5,cEn = sess.run([top1_accuracy,top5_accuracy,loss_operation],
                              feed_dict={images:batch_x, y:batch_y, RC: regConst, KP:1, train_mode:False})
        total_top1 += t1
        total_top5 += t5
    total_crossEn += (cEn *len(batch_x))
    total_top1 /= nImgs
    total_top5 /= nImgs
    total_crossEn /= nImgs
    return total_top1, total_top5, total_crossEn