Note
Click here to download the full example code
Regression tutorial
This tutorial demonstrates that hardware-compatible Akida models can perform regression tasks at the same accuracy level as a native CNN network.
This is illustrated through an age estimation problem using the UTKFace dataset.
1. Load the dataset
from akida_models.utk_face.preprocessing import load_data
# Load the dataset using akida_models preprocessing tool
x_train, y_train, x_test, y_test = load_data()
# For Akida inference, use uint8 raw data
x_test_akida = x_test.astype('uint8')
Out:
Downloading data from http://data.brainchip.com/dataset-mirror/utk_face/UTKFace_preprocessed.tar.gz
16384/48742400 [..............................] - ETA: 1s
139264/48742400 [..............................] - ETA: 22s
335872/48742400 [..............................] - ETA: 17s
532480/48742400 [..............................] - ETA: 16s
729088/48742400 [..............................] - ETA: 15s
925696/48742400 [..............................] - ETA: 15s
1122304/48742400 [..............................] - ETA: 15s
1318912/48742400 [..............................] - ETA: 15s
1515520/48742400 [..............................] - ETA: 14s
1712128/48742400 [>.............................] - ETA: 14s
1908736/48742400 [>.............................] - ETA: 14s
2105344/48742400 [>.............................] - ETA: 14s
2301952/48742400 [>.............................] - ETA: 14s
2498560/48742400 [>.............................] - ETA: 14s
2695168/48742400 [>.............................] - ETA: 14s
2891776/48742400 [>.............................] - ETA: 14s
3088384/48742400 [>.............................] - ETA: 14s
3284992/48742400 [=>............................] - ETA: 14s
3481600/48742400 [=>............................] - ETA: 13s
3678208/48742400 [=>............................] - ETA: 13s
3874816/48742400 [=>............................] - ETA: 13s
4071424/48742400 [=>............................] - ETA: 13s
4268032/48742400 [=>............................] - ETA: 13s
4464640/48742400 [=>............................] - ETA: 13s
4661248/48742400 [=>............................] - ETA: 13s
4849664/48742400 [=>............................] - ETA: 13s
4988928/48742400 [==>...........................] - ETA: 13s
5185536/48742400 [==>...........................] - ETA: 13s
5382144/48742400 [==>...........................] - ETA: 13s
5578752/48742400 [==>...........................] - ETA: 13s
5775360/48742400 [==>...........................] - ETA: 13s
5971968/48742400 [==>...........................] - ETA: 13s
6168576/48742400 [==>...........................] - ETA: 13s
6365184/48742400 [==>...........................] - ETA: 13s
6561792/48742400 [===>..........................] - ETA: 13s
6758400/48742400 [===>..........................] - ETA: 13s
6955008/48742400 [===>..........................] - ETA: 13s
7151616/48742400 [===>..........................] - ETA: 12s
7348224/48742400 [===>..........................] - ETA: 12s
7544832/48742400 [===>..........................] - ETA: 12s
7741440/48742400 [===>..........................] - ETA: 12s
7938048/48742400 [===>..........................] - ETA: 12s
8134656/48742400 [====>.........................] - ETA: 12s
8331264/48742400 [====>.........................] - ETA: 12s
8527872/48742400 [====>.........................] - ETA: 12s
8724480/48742400 [====>.........................] - ETA: 12s
8921088/48742400 [====>.........................] - ETA: 12s
9117696/48742400 [====>.........................] - ETA: 12s
9314304/48742400 [====>.........................] - ETA: 12s
9510912/48742400 [====>.........................] - ETA: 12s
9707520/48742400 [====>.........................] - ETA: 12s
9904128/48742400 [=====>........................] - ETA: 12s
10100736/48742400 [=====>........................] - ETA: 11s
10297344/48742400 [=====>........................] - ETA: 11s
10493952/48742400 [=====>........................] - ETA: 11s
10690560/48742400 [=====>........................] - ETA: 11s
10887168/48742400 [=====>........................] - ETA: 11s
11083776/48742400 [=====>........................] - ETA: 11s
11280384/48742400 [=====>........................] - ETA: 11s
11476992/48742400 [======>.......................] - ETA: 11s
11673600/48742400 [======>.......................] - ETA: 11s
11870208/48742400 [======>.......................] - ETA: 11s
12066816/48742400 [======>.......................] - ETA: 11s
12263424/48742400 [======>.......................] - ETA: 11s
12460032/48742400 [======>.......................] - ETA: 11s
12656640/48742400 [======>.......................] - ETA: 11s
12853248/48742400 [======>.......................] - ETA: 11s
13049856/48742400 [=======>......................] - ETA: 10s
13246464/48742400 [=======>......................] - ETA: 10s
13443072/48742400 [=======>......................] - ETA: 10s
13639680/48742400 [=======>......................] - ETA: 10s
13836288/48742400 [=======>......................] - ETA: 10s
14032896/48742400 [=======>......................] - ETA: 10s
14229504/48742400 [=======>......................] - ETA: 10s
14426112/48742400 [=======>......................] - ETA: 10s
14622720/48742400 [========>.....................] - ETA: 10s
14819328/48742400 [========>.....................] - ETA: 10s
15015936/48742400 [========>.....................] - ETA: 10s
15212544/48742400 [========>.....................] - ETA: 10s
15409152/48742400 [========>.....................] - ETA: 10s
15605760/48742400 [========>.....................] - ETA: 10s
15802368/48742400 [========>.....................] - ETA: 10s
15998976/48742400 [========>.....................] - ETA: 10s
16195584/48742400 [========>.....................] - ETA: 9s
16392192/48742400 [=========>....................] - ETA: 9s
16523264/48742400 [=========>....................] - ETA: 10s
16539648/48742400 [=========>....................] - ETA: 11s
16580608/48742400 [=========>....................] - ETA: 11s
16621568/48742400 [=========>....................] - ETA: 11s
16687104/48742400 [=========>....................] - ETA: 11s
16760832/48742400 [=========>....................] - ETA: 11s
16850944/48742400 [=========>....................] - ETA: 11s
16949248/48742400 [=========>....................] - ETA: 11s
17047552/48742400 [=========>....................] - ETA: 11s
17178624/48742400 [=========>....................] - ETA: 11s
17293312/48742400 [=========>....................] - ETA: 11s
17375232/48742400 [=========>....................] - ETA: 11s
17383424/48742400 [=========>....................] - ETA: 12s
17440768/48742400 [=========>....................] - ETA: 12s
17506304/48742400 [=========>....................] - ETA: 12s
17596416/48742400 [=========>....................] - ETA: 12s
17694720/48742400 [=========>....................] - ETA: 12s
17801216/48742400 [=========>....................] - ETA: 12s
17924096/48742400 [==========>...................] - ETA: 12s
18030592/48742400 [==========>...................] - ETA: 12s
18161664/48742400 [==========>...................] - ETA: 12s
18292736/48742400 [==========>...................] - ETA: 12s
18423808/48742400 [==========>...................] - ETA: 12s
18554880/48742400 [==========>...................] - ETA: 11s
18685952/48742400 [==========>...................] - ETA: 11s
18882560/48742400 [==========>...................] - ETA: 11s
19079168/48742400 [==========>...................] - ETA: 11s
19275776/48742400 [==========>...................] - ETA: 11s
19472384/48742400 [==========>...................] - ETA: 11s
19668992/48742400 [===========>..................] - ETA: 11s
19857408/48742400 [===========>..................] - ETA: 11s
19996672/48742400 [===========>..................] - ETA: 11s
20193280/48742400 [===========>..................] - ETA: 11s
20389888/48742400 [===========>..................] - ETA: 11s
20586496/48742400 [===========>..................] - ETA: 10s
20783104/48742400 [===========>..................] - ETA: 10s
20979712/48742400 [===========>..................] - ETA: 10s
21176320/48742400 [============>.................] - ETA: 10s
21372928/48742400 [============>.................] - ETA: 10s
21569536/48742400 [============>.................] - ETA: 10s
21766144/48742400 [============>.................] - ETA: 10s
21962752/48742400 [============>.................] - ETA: 10s
22159360/48742400 [============>.................] - ETA: 10s
22355968/48742400 [============>.................] - ETA: 10s
22552576/48742400 [============>.................] - ETA: 10s
22675456/48742400 [============>.................] - ETA: 9s
22814720/48742400 [=============>................] - ETA: 9s
23011328/48742400 [=============>................] - ETA: 9s
23207936/48742400 [=============>................] - ETA: 9s
23404544/48742400 [=============>................] - ETA: 9s
23601152/48742400 [=============>................] - ETA: 9s
23797760/48742400 [=============>................] - ETA: 9s
23994368/48742400 [=============>................] - ETA: 9s
24190976/48742400 [=============>................] - ETA: 9s
24387584/48742400 [==============>...............] - ETA: 9s
24584192/48742400 [==============>...............] - ETA: 9s
24780800/48742400 [==============>...............] - ETA: 9s
24977408/48742400 [==============>...............] - ETA: 8s
25100288/48742400 [==============>...............] - ETA: 9s
25149440/48742400 [==============>...............] - ETA: 9s
25165824/48742400 [==============>...............] - ETA: 9s
25198592/48742400 [==============>...............] - ETA: 9s
25247744/48742400 [==============>...............] - ETA: 9s
25305088/48742400 [==============>...............] - ETA: 9s
25395200/48742400 [==============>...............] - ETA: 9s
25493504/48742400 [==============>...............] - ETA: 9s
25526272/48742400 [==============>...............] - ETA: 9s
25600000/48742400 [==============>...............] - ETA: 9s
25640960/48742400 [==============>...............] - ETA: 9s
25690112/48742400 [==============>...............] - ETA: 9s
25747456/48742400 [==============>...............] - ETA: 9s
25829376/48742400 [==============>...............] - ETA: 9s
25919488/48742400 [==============>...............] - ETA: 9s
26017792/48742400 [===============>..............] - ETA: 9s
26124288/48742400 [===============>..............] - ETA: 9s
26247168/48742400 [===============>..............] - ETA: 9s
26353664/48742400 [===============>..............] - ETA: 9s
26484736/48742400 [===============>..............] - ETA: 9s
26615808/48742400 [===============>..............] - ETA: 9s
26746880/48742400 [===============>..............] - ETA: 9s
26877952/48742400 [===============>..............] - ETA: 9s
27009024/48742400 [===============>..............] - ETA: 9s
27205632/48742400 [===============>..............] - ETA: 9s
27402240/48742400 [===============>..............] - ETA: 8s
27598848/48742400 [===============>..............] - ETA: 8s
27795456/48742400 [================>.............] - ETA: 8s
27992064/48742400 [================>.............] - ETA: 8s
28188672/48742400 [================>.............] - ETA: 8s
28385280/48742400 [================>.............] - ETA: 8s
28581888/48742400 [================>.............] - ETA: 8s
28778496/48742400 [================>.............] - ETA: 8s
28975104/48742400 [================>.............] - ETA: 8s
29171712/48742400 [================>.............] - ETA: 8s
29368320/48742400 [=================>............] - ETA: 8s
29499392/48742400 [=================>............] - ETA: 7s
29507584/48742400 [=================>............] - ETA: 8s
29564928/48742400 [=================>............] - ETA: 8s
29630464/48742400 [=================>............] - ETA: 8s
29638656/48742400 [=================>............] - ETA: 8s
29679616/48742400 [=================>............] - ETA: 8s
29745152/48742400 [=================>............] - ETA: 8s
29794304/48742400 [=================>............] - ETA: 8s
29827072/48742400 [=================>............] - ETA: 8s
29851648/48742400 [=================>............] - ETA: 8s
29884416/48742400 [=================>............] - ETA: 8s
29933568/48742400 [=================>............] - ETA: 8s
29990912/48742400 [=================>............] - ETA: 8s
30064640/48742400 [=================>............] - ETA: 8s
30154752/48742400 [=================>............] - ETA: 8s
30253056/48742400 [=================>............] - ETA: 8s
30384128/48742400 [=================>............] - ETA: 8s
30515200/48742400 [=================>............] - ETA: 8s
30613504/48742400 [=================>............] - ETA: 8s
30744576/48742400 [=================>............] - ETA: 8s
30875648/48742400 [==================>...........] - ETA: 8s
31006720/48742400 [==================>...........] - ETA: 8s
31137792/48742400 [==================>...........] - ETA: 8s
31285248/48742400 [==================>...........] - ETA: 7s
31465472/48742400 [==================>...........] - ETA: 7s
31662080/48742400 [==================>...........] - ETA: 7s
31858688/48742400 [==================>...........] - ETA: 7s
32055296/48742400 [==================>...........] - ETA: 7s
32251904/48742400 [==================>...........] - ETA: 7s
32448512/48742400 [==================>...........] - ETA: 7s
32645120/48742400 [===================>..........] - ETA: 7s
32841728/48742400 [===================>..........] - ETA: 7s
33038336/48742400 [===================>..........] - ETA: 7s
33234944/48742400 [===================>..........] - ETA: 6s
33431552/48742400 [===================>..........] - ETA: 6s
33628160/48742400 [===================>..........] - ETA: 6s
33824768/48742400 [===================>..........] - ETA: 6s
34021376/48742400 [===================>..........] - ETA: 6s
34217984/48742400 [====================>.........] - ETA: 6s
34414592/48742400 [====================>.........] - ETA: 6s
34611200/48742400 [====================>.........] - ETA: 6s
34807808/48742400 [====================>.........] - ETA: 6s
35004416/48742400 [====================>.........] - ETA: 6s
35201024/48742400 [====================>.........] - ETA: 5s
35397632/48742400 [====================>.........] - ETA: 5s
35594240/48742400 [====================>.........] - ETA: 5s
35790848/48742400 [=====================>........] - ETA: 5s
35987456/48742400 [=====================>........] - ETA: 5s
36184064/48742400 [=====================>........] - ETA: 5s
36380672/48742400 [=====================>........] - ETA: 5s
36577280/48742400 [=====================>........] - ETA: 5s
36773888/48742400 [=====================>........] - ETA: 5s
36970496/48742400 [=====================>........] - ETA: 5s
37167104/48742400 [=====================>........] - ETA: 5s
37363712/48742400 [=====================>........] - ETA: 4s
37560320/48742400 [======================>.......] - ETA: 4s
37756928/48742400 [======================>.......] - ETA: 4s
37953536/48742400 [======================>.......] - ETA: 4s
38150144/48742400 [======================>.......] - ETA: 4s
38346752/48742400 [======================>.......] - ETA: 4s
38543360/48742400 [======================>.......] - ETA: 4s
38739968/48742400 [======================>.......] - ETA: 4s
38936576/48742400 [======================>.......] - ETA: 4s
39133184/48742400 [=======================>......] - ETA: 4s
39329792/48742400 [=======================>......] - ETA: 3s
39526400/48742400 [=======================>......] - ETA: 3s
39723008/48742400 [=======================>......] - ETA: 3s
39919616/48742400 [=======================>......] - ETA: 3s
40116224/48742400 [=======================>......] - ETA: 3s
40312832/48742400 [=======================>......] - ETA: 3s
40509440/48742400 [=======================>......] - ETA: 3s
40706048/48742400 [========================>.....] - ETA: 3s
40902656/48742400 [========================>.....] - ETA: 3s
41099264/48742400 [========================>.....] - ETA: 3s
41295872/48742400 [========================>.....] - ETA: 3s
41492480/48742400 [========================>.....] - ETA: 3s
41689088/48742400 [========================>.....] - ETA: 2s
41885696/48742400 [========================>.....] - ETA: 2s
42082304/48742400 [========================>.....] - ETA: 2s
42278912/48742400 [=========================>....] - ETA: 2s
42475520/48742400 [=========================>....] - ETA: 2s
42672128/48742400 [=========================>....] - ETA: 2s
42868736/48742400 [=========================>....] - ETA: 2s
43065344/48742400 [=========================>....] - ETA: 2s
43261952/48742400 [=========================>....] - ETA: 2s
43458560/48742400 [=========================>....] - ETA: 2s
43655168/48742400 [=========================>....] - ETA: 2s
43851776/48742400 [=========================>....] - ETA: 2s
44048384/48742400 [==========================>...] - ETA: 1s
44244992/48742400 [==========================>...] - ETA: 1s
44441600/48742400 [==========================>...] - ETA: 1s
44638208/48742400 [==========================>...] - ETA: 1s
44834816/48742400 [==========================>...] - ETA: 1s
45031424/48742400 [==========================>...] - ETA: 1s
45228032/48742400 [==========================>...] - ETA: 1s
45424640/48742400 [==========================>...] - ETA: 1s
45621248/48742400 [===========================>..] - ETA: 1s
45817856/48742400 [===========================>..] - ETA: 1s
46014464/48742400 [===========================>..] - ETA: 1s
46211072/48742400 [===========================>..] - ETA: 1s
46407680/48742400 [===========================>..] - ETA: 0s
46604288/48742400 [===========================>..] - ETA: 0s
46800896/48742400 [===========================>..] - ETA: 0s
46997504/48742400 [===========================>..] - ETA: 0s
47194112/48742400 [============================>.] - ETA: 0s
47390720/48742400 [============================>.] - ETA: 0s
47587328/48742400 [============================>.] - ETA: 0s
47783936/48742400 [============================>.] - ETA: 0s
47980544/48742400 [============================>.] - ETA: 0s
48177152/48742400 [============================>.] - ETA: 0s
48373760/48742400 [============================>.] - ETA: 0s
48570368/48742400 [============================>.] - ETA: 0s
48742400/48742400 [==============================] - 20s 0us/step
48750592/48742400 [==============================] - 20s 0us/step
2. Load a pre-trained native Keras model
The model is a simplified version inspired from VGG architecture. It consists of a succession of convolutional and pooling layers and ends with two fully connected layers that outputs a single value corresponding to the estimated age. This model architecture is compatible with the design constraints before quantization. It is the starting point for a model runnable on the Akida NSoC.
The pre-trained native Keras model loaded below was trained on 300 epochs. The model file is available on the BrainChip data server.
The performance of the model is evaluated using the “Mean Absolute Error” (MAE). The MAE, used as a metric in regression problem, is calculated as an average of absolute differences between the target values and the predictions. The MAE is a linear score, i.e. all the individual differences are equally weighted in the average.
from tensorflow.keras.utils import get_file
from tensorflow.keras.models import load_model
# Retrieve the model file from the BrainChip data server
model_file = get_file("vgg_utk_face.h5",
"http://data.brainchip.com/models/vgg/vgg_utk_face.h5",
cache_subdir='models')
# Load the native Keras pre-trained model
model_keras = load_model(model_file)
model_keras.summary()
Out:
Downloading data from http://data.brainchip.com/models/vgg/vgg_utk_face.h5
16384/1907648 [..............................] - ETA: 0s
139264/1907648 [=>............................] - ETA: 0s
335872/1907648 [====>.........................] - ETA: 0s
532480/1907648 [=======>......................] - ETA: 0s
729088/1907648 [==========>...................] - ETA: 0s
925696/1907648 [=============>................] - ETA: 0s
1122304/1907648 [================>.............] - ETA: 0s
1318912/1907648 [===================>..........] - ETA: 0s
1515520/1907648 [======================>.......] - ETA: 0s
1712128/1907648 [=========================>....] - ETA: 0s
1908736/1907648 [==============================] - 1s 0us/step
1916928/1907648 [==============================] - 1s 0us/step
Model: "vgg_utk_face"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_37 (InputLayer) [(None, 32, 32, 3)] 0
rescaling_36 (Rescaling) (None, 32, 32, 3) 0
conv_0 (Conv2D) (None, 30, 30, 32) 864
conv_0_BN (BatchNormalizati (None, 30, 30, 32) 128
on)
conv_0_relu (ReLU) (None, 30, 30, 32) 0
conv_1 (Conv2D) (None, 30, 30, 32) 9216
conv_1_maxpool (MaxPooling2 (None, 15, 15, 32) 0
D)
conv_1_BN (BatchNormalizati (None, 15, 15, 32) 128
on)
conv_1_relu (ReLU) (None, 15, 15, 32) 0
dropout (Dropout) (None, 15, 15, 32) 0
conv_2 (Conv2D) (None, 15, 15, 64) 18432
conv_2_BN (BatchNormalizati (None, 15, 15, 64) 256
on)
conv_2_relu (ReLU) (None, 15, 15, 64) 0
conv_3 (Conv2D) (None, 15, 15, 64) 36864
conv_3_maxpool (MaxPooling2 (None, 8, 8, 64) 0
D)
conv_3_BN (BatchNormalizati (None, 8, 8, 64) 256
on)
conv_3_relu (ReLU) (None, 8, 8, 64) 0
dropout_1 (Dropout) (None, 8, 8, 64) 0
conv_4 (Conv2D) (None, 8, 8, 84) 48384
conv_4_BN (BatchNormalizati (None, 8, 8, 84) 336
on)
conv_4_relu (ReLU) (None, 8, 8, 84) 0
dropout_2 (Dropout) (None, 8, 8, 84) 0
flatten_3 (Flatten) (None, 5376) 0
dense_1 (Dense) (None, 64) 344064
dense_1_BN (BatchNormalizat (None, 64) 256
ion)
dense_1_relu (ReLU) (None, 64) 0
dense_2 (Dense) (None, 1) 65
=================================================================
Total params: 459,249
Trainable params: 458,569
Non-trainable params: 680
_________________________________________________________________
# Compile the native Keras model (required to evaluate the MAE)
model_keras.compile(optimizer='Adam', loss='mae')
# Check Keras model performance
mae_keras = model_keras.evaluate(x_test, y_test, verbose=0)
print("Keras MAE: {0:.4f}".format(mae_keras))
Out:
Keras MAE: 5.8023
3. Load a pre-trained quantized Keras model satisfying Akida NSoC requirements
The above native Keras model is quantized and fine-tuned to get a quantized Keras model satisfying the Akida NSoC requirements. The first convolutional layer of our model uses 8-bit weights and other layers are quantized using 2-bit weights. All activations are 2 bits.
The pre-trained model was obtained after two fine-tuning episodes:
the model is first quantized and fine-tuned with 4-bit weights and activations (first convolutional weights are 8 bits)
the model is then quantized and fine-tuned with 2-bit weights and activations (first convolutional weights are still 8 bits).
The table below summarizes the “Mean Absolute Error” (MAE) results obtained after every training episode.
Episode |
Weights Quant. |
Activ. Quant. |
MAE |
Epochs |
|---|---|---|---|---|
1 |
N/A |
N/A |
5.80 |
300 |
2 |
8/4 bits |
4 bits |
5.79 |
30 |
3 |
8/2 bits |
2 bits |
6.15 |
30 |
Here, we directly load the pre-trained quantized Keras model using the akida_models helper.
from akida_models import vgg_utk_face_pretrained
# Load the pre-trained quantized model
model_quantized_keras = vgg_utk_face_pretrained()
model_quantized_keras.summary()
Out:
Downloading data from http://data.brainchip.com/models/vgg/vgg_utk_face_iq8_wq2_aq2.h5
16384/1877128 [..............................] - ETA: 0s
139264/1877128 [=>............................] - ETA: 0s
335872/1877128 [====>.........................] - ETA: 0s
532480/1877128 [=======>......................] - ETA: 0s
729088/1877128 [==========>...................] - ETA: 0s
925696/1877128 [=============>................] - ETA: 0s
1122304/1877128 [================>.............] - ETA: 0s
1318912/1877128 [====================>.........] - ETA: 0s
1515520/1877128 [=======================>......] - ETA: 0s
1712128/1877128 [==========================>...] - ETA: 0s
1884160/1877128 [==============================] - 1s 0us/step
1892352/1877128 [==============================] - 1s 0us/step
Model: "vgg_utk_face"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_38 (InputLayer) [(None, 32, 32, 3)] 0
rescaling_37 (Rescaling) (None, 32, 32, 3) 0
conv_0 (QuantizedConv2D) (None, 30, 30, 32) 896
conv_0_relu (ActivationDisc (None, 30, 30, 32) 0
reteRelu)
conv_1 (QuantizedConv2D) (None, 30, 30, 32) 9248
conv_1_maxpool (MaxPooling2 (None, 15, 15, 32) 0
D)
conv_1_relu (ActivationDisc (None, 15, 15, 32) 0
reteRelu)
dropout_3 (Dropout) (None, 15, 15, 32) 0
conv_2 (QuantizedConv2D) (None, 15, 15, 64) 18496
conv_2_relu (ActivationDisc (None, 15, 15, 64) 0
reteRelu)
conv_3 (QuantizedConv2D) (None, 15, 15, 64) 36928
conv_3_maxpool (MaxPooling2 (None, 8, 8, 64) 0
D)
conv_3_relu (ActivationDisc (None, 8, 8, 64) 0
reteRelu)
dropout_4 (Dropout) (None, 8, 8, 64) 0
conv_4 (QuantizedConv2D) (None, 8, 8, 84) 48468
conv_4_relu (ActivationDisc (None, 8, 8, 84) 0
reteRelu)
dropout_5 (Dropout) (None, 8, 8, 84) 0
flatten_4 (Flatten) (None, 5376) 0
dense_1 (QuantizedDense) (None, 64) 344128
dense_1_relu (ActivationDis (None, 64) 0
creteRelu)
dense_2 (QuantizedDense) (None, 1) 65
=================================================================
Total params: 458,229
Trainable params: 458,229
Non-trainable params: 0
_________________________________________________________________
# Compile the quantized Keras model (required to evaluate the MAE)
model_quantized_keras.compile(optimizer='Adam', loss='mae')
# Check Keras model performance
mae_quant = model_quantized_keras.evaluate(x_test, y_test, verbose=0)
print("Keras MAE: {0:.4f}".format(mae_quant))
Out:
Keras MAE: 6.1465
4. Conversion to Akida
The quantized Keras model is now converted into an Akida model. After conversion, we evaluate the performance on the UTKFace dataset.
Since activations sparsity has a great impact on Akida inference time, we also have a look at the average input and output sparsity of each layer on a subset of the dataset.
from cnn2snn import convert
# Convert the model
model_akida = convert(model_quantized_keras)
model_akida.summary()
Out:
Model Summary
______________________________________________
Input shape Output shape Sequences Layers
==============================================
[32, 32, 3] [1, 1, 1] 1 7
______________________________________________
SW/conv_0-dense_2 (Software)
_____________________________________________________
Layer (type) Output shape Kernel shape
=====================================================
conv_0 (InputConv.) [30, 30, 32] (3, 3, 3, 32)
_____________________________________________________
conv_1 (Conv.) [15, 15, 32] (3, 3, 32, 32)
_____________________________________________________
conv_2 (Conv.) [15, 15, 64] (3, 3, 32, 64)
_____________________________________________________
conv_3 (Conv.) [8, 8, 64] (3, 3, 64, 64)
_____________________________________________________
conv_4 (Conv.) [8, 8, 84] (3, 3, 64, 84)
_____________________________________________________
dense_1 (Fully.) [1, 1, 64] (1, 1, 5376, 64)
_____________________________________________________
dense_2 (Fully.) [1, 1, 1] (1, 1, 64, 1)
_____________________________________________________
import numpy as np
# Check Akida model performance
y_akida = model_akida.predict(x_test_akida)
# Compute and display the MAE
mae_akida = np.sum(np.abs(y_test.squeeze() - y_akida.squeeze())) / len(y_test)
print("Akida MAE: {0:.4f}".format(mae_akida))
# For non-regression purpose
assert abs(mae_keras - mae_akida) < 0.5
Out:
Akida MAE: 6.1791
Let’s summarize the MAE performance for the native Keras, the quantized Keras and the Akida model.
Model |
MAE |
|---|---|
native Keras |
5.80 |
quantized Keras |
6.15 |
Akida |
6.21 |
5. Estimate age on a single image
import matplotlib.pyplot as plt
# Estimate age on a random single image and display Keras and Akida outputs
id = np.random.randint(0, len(y_test) + 1)
age_keras = model_keras.predict(x_test[id:id + 1])
plt.imshow(x_test_akida[id], interpolation='bicubic')
plt.xticks([]), plt.yticks([])
plt.show()
print("Keras estimated age: {0:.1f}".format(age_keras.squeeze()))
print("Akida estimated age: {0:.1f}".format(y_akida[id].squeeze()))
print(f"Actual age: {y_test[id].squeeze()}")
Out:
Keras estimated age: 26.1
Akida estimated age: 25.8
Actual age: 29
Total running time of the script: ( 0 minutes 36.902 seconds)