Sane.AI.Spectrogram v3 - Impulses

Experiments (2 / 10)

	Name	Input	DSP Blocks	Learn Blocks	f32_V_Acc	f32_T_Acc	i8_V_Acc	i8_T_Acc	Tr_Proc	f32_Latency	f32_RAM	f32_Flash	i8_Latency	i8_RAM	i8_Flash
	spectr-conv1d-123	2,000ms. @ 16000Hz	Spectrogram	Classifier	54.8%	0.0%	54.8%	-	CPU	168 ms.	106.0K	107.3K	162 ms.	106.0K	51.8K	View impulse
	spectrogram	5,000ms. @ 48000Hz	Spectral features , MFE	Classifier, Classifier via features	100.0%	86.4%	100.0%	86.4%	CPU	4 ms.	32.9K	539.6K	2 ms.	11.2K	170.7K	View impulse

Impulse information

Tags - Impulse tags

Source - Impulse source (Manual vs EON Tuner)

Created by - The user who originally created this impulse

Key metrics

f32_V_Acc - Accuracy (float32, validation set)

f32_T_Acc - Accuracy (float32, test set)

i8_V_Acc - Accuracy (int8, validation set)

i8_T_Acc - Accuracy (int8, test set)

Tr_Proc - Training processor

f32_Latency - float32 Latency for the complete impulse.

f32_RAM - float32 RAM for the complete impulse.

f32_Flash - float32 Flash for the complete impulse.

i8_Latency - int8 Latency for the complete impulse.

i8_RAM - int8 RAM for the complete impulse.

i8_Flash - int8 Flash for the complete impulse.

Input block config

Win_Size - Window size

Win_Incr - Window increase

Freq - Frequency

Zero_Pad - Zero pad data

Processing block config

P_1_Norm - Normalization

P_1_Fra_len - Spectrogram > Frame length (Processing Block #1)

P_1_Fra_str - Spectrogram > Frame stride (Processing Block #1)

P_1_FFT_len - Spectrogram > FFT length (Processing Block #1)

P_1_Noi_flo - Normalization > Noise floor (dB) (Processing Block #1)

P_1_Sca_axe - Filter > Scale axes (Processing Block #1)

P_1_Idr - Filter > Input decimation ratio (Processing Block #1)

P_1_Typ - Filter > Type (Processing Block #1)

P_1_Cof - Filter > Cut-off frequency (Processing Block #1)

P_1_Ord - Filter > Order (Processing Block #1)

P_1_Typ_ATPB1 - Analysis > Type (Processing Block #1)

P_1_Tlos - Analysis > Take log of spectrum? (Processing Block #1)

P_1_OFf - Analysis > Overlap FFT frames? (Processing Block #1)

P_1_WDL - Analysis > Wavelet Decomposition Level (Processing Block #1)

P_1_Wav - Analysis > Wavelet (Processing Block #1)

P_1_Ilfr - Analysis > Improve low frequency resolution? (Processing Block #1)

P_2_Norm - Normalization

P_2_Fra_len - Mel-filterbank energy features > Frame length (Processing Block #2)

P_2_Fra_str - Mel-filterbank energy features > Frame stride (Processing Block #2)

P_2_Fil_num - Mel-filterbank energy features > Filter number (Processing Block #2)

P_2_FFT_len - Mel-filterbank energy features > FFT length (Processing Block #2)

P_2_Low_fre - Mel-filterbank energy features > Low frequency (Processing Block #2)

P_2_Hig_fre - Mel-filterbank energy features > High frequency (Processing Block #2)

P_2_Noi_flo - Normalization > Noise floor (dB) (Processing Block #2)

Learn block config

L_1_Model - ML Model architecture (Learn Block #1)

L_1_Tc - Training cycles (Learn Block #1)

L_1_Lr - Learning rate (Learn Block #1)

L_1_Lo - Use learned optimizer? (Learn Block #1)

L_1_Bs - Batch size (Learn Block #1)

L_1_Acw - Auto-weight classes (Learn Block #1)

L_1_Data_Aug - Data augmentation (Learn Block #1)

L_1_Val_Size - Validation set size (Learn Block #1)

L_1_Val_Key - Split train/validation set on metadata key (Learn Block #1)

L_1_Mode - Visual / expert mode (Learn Block #1)

L_1_Thr_Mcr - Minimum confidence rating (Learn Block #1)

L_2_Model - ML Model architecture (Learn Block #2)

L_2_Tc - Training cycles (Learn Block #2)

L_2_Lr - Learning rate (Learn Block #2)

L_2_Lo - Use learned optimizer? (Learn Block #2)

L_2_Acw - Auto-weight classes (Learn Block #2)

L_2_Data_Aug - Data augmentation (Learn Block #2)

L_2_Val_Size - Validation set size (Learn Block #2)

L_2_Val_Key - Split train/validation set on metadata key (Learn Block #2)

L_2_Mode - Visual / expert mode (Learn Block #2)

L_2_Thr_Mcr - Minimum confidence rating (Learn Block #2)

Accuracy metrics

f32_V_AUC - float32 Area under ROC Curve (validation set)

f32_V_WA_P - float32 Weighted average Precision (validation set)

f32_V_WA_R - float32 Weighted average Recall (validation set)

f32_V_WA_F1 - float32 Weighted average F1 score (validation set)

f32_T_AUC - float32 Area under ROC Curve (test set)

f32_T_WA_P - float32 Weighted average Precision (test set)

f32_T_WA_R - float32 Weighted average Recall (test set)

f32_T_WA_F1 - float32 Weighted average F1 score (test set)

i8_V_AUC - int8 Area under ROC Curve (validation set)

i8_V_WA_P - int8 Weighted average Precision (validation set)

i8_V_WA_R - int8 Weighted average Recall (validation set)

i8_V_WA_F1 - int8 Weighted average F1 score (validation set)

i8_T_AUC - int8 Area under ROC Curve (test set)

i8_T_WA_P - int8 Weighted average Precision (test set)

i8_T_WA_R - int8 Weighted average Recall (test set)

i8_T_WA_F1 - int8 Weighted average F1 score (test set)

Post-processing block config

No columns

Target device

Define your target device requirements to inform model optimizations and performance calculations. No device yet? Use the default settings which you can change at any time.

Target device

Processor family

Clock rate

| MHz

Max

Custom device name

Application budget

Specify the available RAM and ROM for the model's operation, along with the maximum allowed latency for your specific application. Not sure yet? Start with the defaults and modify them later on.

RAM

| KB

Max

ROM

| KB

Max

Latency

| ms

Max

Experiments (2 / 10)

spectr-conv1d-123

spectrogram