remove unused Performance section from models .MD files

models/intel/action-recognition-0001/description/action-recognition-0001.md

@@ -27,8 +27,6 @@ Video frames should be sampled to cover ~1 second fragment (i.e. skip every seco
 | MParams                         | 21.276                                    |
 
 
-### Performance
-
 ### Inputs
 
 1. name: "0", shape: [1x3x224x224] - An input image in the format [BxCxHxW],
@@ -55,8 +53,6 @@ The action-recognition-0001-decoder model accepts stack of frame embeddings, com
 | MParams                         | 4.405                                     |
 
 
-### Performance
-
 ### Inputs
 
 1. name: "0" , shape: [1x16x512] - An embedding image in the format [BxTxC],

models/intel/age-gender-recognition-retail-0013/description/age-gender-recognition-retail-0013.md

@@ -36,8 +36,6 @@ applicable for children since their faces were not in the training set.
 | Avg. age error  | 6.99 years |
 | Gender accuracy |     95.80% |
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x62x62] - An input image in [1xCxHxW] format. Expected color order is BGR.

models/intel/asl-recognition-0004/description/asl-recognition-0004.md

@@ -23,8 +23,6 @@ on the input clip.
 | MParams                         | 4.133                                     |
 | Source framework                | PyTorch\*                                 |
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x16x224x224]. An input image sequence in the format [BxCxTxHxW], where:

models/intel/bert-large-uncased-whole-word-masking-squad-emb-0001/description/bert-large-uncased-whole-word-masking-squad-emb-0001.md

@@ -36,8 +36,6 @@ The top5 is calculated as follow:
 2. For each question from  SQuAD v1.1 dev set the question embedding vector is calculated and compared with each previously calculated context embedding vector. If the right context is in top 5 context embedding closest to question embedding then top5_count increased by 1.
 3. top5 = top5_count / question_number
 
-## Performance
-
 ## Input
 
 1. Token IDs, name: `input_ids`, shape: [1x384] for context and [1x32] for question.

models/intel/bert-large-uncased-whole-word-masking-squad-fp32-0001/description/bert-large-uncased-whole-word-masking-squad-fp32-0001.md

@@ -25,8 +25,6 @@ The quality metrics were calculated on the SQuAD v1.1 dataset ("dev" split). Max
 | Exact match (EM)          |        87.20% |
 
 
-## Performance
-
 ## Input
 
 Input 0: A `1,384` sequence of tokens (integer values) representing the tokenized premise and question ("input_ids"). The sequence structure is as follows (`[CLS]`, `[SEP]` and `[PAD]` should be replaced by corresponding token IDs as specified by the dictionary):

models/intel/bert-large-uncased-whole-word-masking-squad-int8-0001/description/bert-large-uncased-whole-word-masking-squad-int8-0001.md

@@ -27,8 +27,6 @@ The quality metrics were calculated on the SQuAD v1.1 dataset ("dev" split). Max
 | Exact match (EM)          |        86.36% |
 
 
-## Performance
-
 ## Input
 
 Input 0: A `1,384` sequence of tokens (integer values) representing the tokenized premise and question ("input_ids"). The sequence structure is as follows (`[CLS]`, `[SEP]` and `[PAD]` should be replaced by corresponding token IDs as specified by the dictionary):

models/intel/bert-small-uncased-whole-word-masking-squad-0001/description/bert-small-uncased-whole-word-masking-squad-0001.md

@@ -30,8 +30,6 @@ The quality metrics were calculated on the SQuAD v1.1 dataset ("dev" split). Max
 | F1                        |        91.57% |
 | Exact match (EM)          |        85.04% |
 
-## Performance
-
 ## Input
 
 1. Token IDs, name: `input_ids`, shape: [1x384].

models/intel/bert-small-uncased-whole-word-masking-squad-0002/description/bert-small-uncased-whole-word-masking-squad-0002.md

@@ -30,8 +30,6 @@ The quality metrics were calculated on the SQuAD v1.1 dataset ("dev" split). Max
 | F1                        |        91.9%  |
 | Exact match (EM)          |        85.4%  |
 
-## Performance
-
 ## Input
 
 1. Token IDs, name: `input_ids`, shape: [1x384].

models/intel/driver-action-recognition-adas-0002/description/driver-action-recognition-adas-0002.md

@@ -26,8 +26,6 @@ Video frames should be sampled to cover ~1 second fragment (i.e. skip every seco
 | MParams                         | 2.863                                     |
 
 
-### Performance
-
 ### Inputs
 
 1. name: "0" , shape: [1x3x224x224] - An input image in the format [BxCxHxW],
@@ -54,8 +52,6 @@ The driver-action-recognition-adas-0002-decoder model accepts stack of frame emb
 | MParams                         | 4.205                                     |
 
 
-### Performance
-
 ### Inputs
 
 1. name: "0" , shape: [1x16x512] - An embedding image in the format [BxTxC],

models/intel/emotions-recognition-retail-0003/description/emotions-recognition-retail-0003.md

@@ -34,8 +34,6 @@ only the images containing five aforementioned emotions is chosen. The total amo
 |-----------------|------------|
 | Accuracy        |     70.20% |
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x64x64] - An input image in [1xCxHxW] format. Expected color order is BGR.

models/intel/face-detection-0200/description/face-detection-0200.md

@@ -24,8 +24,6 @@ Average Precision (AP) is defined as an area under the
 curve. All numbers were evaluated by taking into account only faces bigger than
 64 x 64 pixels.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x256x256] - An input image in the format [BxCxHxW],

models/intel/face-detection-0202/description/face-detection-0202.md

@@ -24,8 +24,6 @@ Average Precision (AP) is defined as an area under the
 curve. All numbers were evaluated by taking into account only faces bigger than
 64 x 64 pixels.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x384x384] - An input image in the format [BxCxHxW],

models/intel/face-detection-0204/description/face-detection-0204.md

@@ -24,8 +24,6 @@ Average Precision (AP) is defined as an area under the
 curve. All numbers were evaluated by taking into account only faces bigger than
 64 x 64 pixels.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x448x448] - An input image in the format [BxCxHxW],

models/intel/face-detection-0205/description/face-detection-0205.md

@@ -23,8 +23,6 @@ Average Precision (AP) is defined as an area under the
 curve. All numbers were evaluated by taking into account only faces bigger than
 64 x 64 pixels.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x416x416] - An input image in the format [BxCxHxW],

models/intel/face-detection-0206/description/face-detection-0206.md

@@ -23,8 +23,6 @@ Average Precision (AP) is defined as an area under the
 curve. All numbers were evaluated by taking into account only faces bigger than
 64 x 64 pixels.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x640x640] - An input image in the format [BxCxHxW],

models/intel/face-detection-adas-0001/description/face-detection-adas-0001.md

@@ -28,8 +28,6 @@ Average Precision (AP) is defined as an area under the
 curve. Numbers are on
 [Wider Face](http://shuoyang1213.me/WIDERFACE/) validation subset.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x384x672] - An input image in the format [BxCxHxW],

models/intel/face-detection-retail-0004/description/face-detection-retail-0004.md

@@ -25,8 +25,6 @@ Average Precision (AP) is defined as an area under the
 curve. All numbers were evaluated by taking into account only faces bigger than
 60 x 60 pixels.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x300x300] - An input image in the format [BxCxHxW],

models/intel/face-detection-retail-0005/description/face-detection-retail-0005.md

@@ -24,8 +24,6 @@ Average Precision (AP) is defined as an area under the
 curve. All numbers were evaluated by taking into account only faces bigger than
 60 x 60 pixels.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x300x300] - An input image in the format [BxCxHxW],

models/intel/facial-landmarks-35-adas-0002/description/facial-landmarks-35-adas-0002.md

@@ -60,8 +60,6 @@ where N is the number of landmarks, _p_-hat and _p_ are, correspondingly, the pr
 | Internal dataset | 0.106 | 0.143 | 0.038 |
 
 
-## Performance
-
 ## Inputs
 
 * Blob in the format [BxCxHxW]

models/intel/faster-rcnn-resnet101-coco-sparse-60-0001/description/faster-rcnn-resnet101-coco-sparse-60-0001.md

@@ -19,8 +19,6 @@ The model input is a blob that consists of a single image of `1x3x800x1280` in t
 
 See Average Precision metric description at [COCO: Common Objects in Context](https://cocodataset.org/#detection-eval). The primary challenge metric is used. Tested on the COCO validation dataset.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x800x1280] - An input image in the format [BxCxHxW],

models/intel/formula-recognition-medium-scan-0001/description/formula-recognition-medium-scan-0001.md

@@ -36,8 +36,6 @@ The formula-recognition-medium-scan-0001-encoder model is a ResNeXt-50 like back
 | MParams                                       | 1.69      |
 
 
-### Performance
-
 ### Inputs
 
 1.	Name: `imgs` , shape: [1x3x160x1400]. An input image in the [1xCxHxW] format.
@@ -62,8 +60,6 @@ The formula-recognition-medium-scan-0001-decoder model is an LSTM based decoder
 
 
 
-### Performance
-
 ### Inputs
 
 1.	Name: `dec_st_c` , shape: [1x512]. Current context state of the LSTM cell.

models/intel/formula-recognition-polynomials-handwritten-0001/description/formula-recognition-polynomials-handwritten-0001.md

@@ -37,8 +37,6 @@ The formula-recognition-polynomials-handwritten-0001-encoder model is a ResNeXt-
 | MParams                                       | 8.6838    |
 
 
-### Performance
-
 ### Inputs
 
 1.	Name: `imgs` , shape: [1x3x96x990]. An input image in the [1xCxHxW] format.
@@ -63,8 +61,6 @@ The formula-recognition-polynomials-handwritten-0001-decoder model is an LSTM ba
 
 
 
-### Performance
-
 ### Inputs
 
 1.	Name: `dec_st_c` , shape: [1x512]. Current context state of the LSTM cell.

models/intel/gaze-estimation-adas-0002/description/gaze-estimation-adas-0002.md

@@ -33,8 +33,6 @@ The accuracy of gaze direction prediction is evaluated through the use of [MAE](
 | Internal dataset | 6.95 | 3.58 |
 
 
-## Performance
-
 ## Inputs
 
 * Blob in the format [BxCxHxW]

models/intel/handwritten-score-recognition-0003/description/handwritten-score-recognition-0003.md

@@ -20,8 +20,6 @@ The network is able to recognize school marks that should have format either `<d
 | MParams                                        | 5.555              |
 | Source framework                               | TensorFlow         |
 
-## Performance
-
 ## Inputs
 
 Shape: [1x1x32x64] - An input image in the format [BxCxHxW],

models/intel/head-pose-estimation-adas-0001/description/head-pose-estimation-adas-0001.md

@@ -31,8 +31,6 @@ one output.
 | pitch |  5.5 ± 5.3                                                                                                                                            |
 | roll  |  4.6 ± 5.6                                                                                                                                            |
 
-## Performance
-
 ## Inputs
 
 1. name: "data" , shape: [1x3x60x60] - An input image in [1xCxHxW] format. Expected color order is BGR.

models/intel/horizontal-text-detection-0001/description/horizontal-text-detection-0001.md

@@ -19,8 +19,6 @@ The key benefit of this model compared to the [base model](../../text-detection-
 | Source framework                                              | PyTorch\*              |
 
 
-## Performance
-
 ## Inputs
 
 1. Name: `input`, shape: [1x3x704x704] - An input image in the format [1xCxHxW],

models/intel/human-pose-estimation-0001/description/human-pose-estimation-0001.md

@@ -23,8 +23,6 @@ Average Precision metric described in [COCO Keypoint Evaluation site](https://co
 
 Tested on a COCO validation subset from the original paper [Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields](https://arxiv.org/abs/1611.08050).
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x256x456]. An input image in the [BxCxHxW] format ,

models/intel/icnet-camvid-ava-0001/description/icnet-camvid-ava-0001.md

@@ -30,8 +30,6 @@ The quality metrics were calculated on the CamVid\* validation dataset. The `unl
   - `FP` - number of false positive pixels for given class
 
 
-## Performance
-
 ## Input
 
 Image, shape - `1,3,720,960`, format is `B,C,H,W` where:

models/intel/icnet-camvid-ava-sparse-30-0001/description/icnet-camvid-ava-sparse-30-0001.md

@@ -30,8 +30,6 @@ The quality metrics were calculated on the CamVid\* validation dataset. The `unl
   - `FP` - number of false positive pixels for given class
 
 
-## Performance
-
 ## Input
 
 Image, shape - `1,3,720,960`, format is `B,C,H,W` where:

models/intel/icnet-camvid-ava-sparse-60-0001/description/icnet-camvid-ava-sparse-60-0001.md

@@ -30,8 +30,6 @@ The quality metrics were calculated on the CamVid\* validation dataset. The `unl
   - `FP` - number of false positive pixels for given class
 
 
-## Performance
-
 ## Input
 
 Image, shape - `1,3,720,960`, format is `B,C,H,W` where:

models/intel/image-retrieval-0001/description/image-retrieval-0001.md

@@ -17,8 +17,6 @@ Image retrieval model based on [MobileNetV2](https://arxiv.org/abs/1801.04381) a
 | MParams                                                       | 2.535                   |
 | Source framework                                              | TensorFlow\*            |
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x224x224] — An input image in the format [BxCxHxW],

models/intel/instance-segmentation-security-0010/description/instance-segmentation-security-0010.md

@@ -27,8 +27,6 @@ in all ROI-wise heads.
 Average Precision (AP) is defined and measured according to standard
 [MS COCO evaluation procedure](https://cocodataset.org/#detection-eval).
 
-## Performance
-
 ## Inputs
 
 1.	name: `im_data` , shape: [1x3x800x1344] - An input image in the format

models/intel/instance-segmentation-security-0050/description/instance-segmentation-security-0050.md

@@ -24,8 +24,6 @@ and light-weight RPN.
 Average Precision (AP) is defined and measured according to standard
 [MS COCO evaluation procedure](https://cocodataset.org/#detection-eval).
 
-## Performance
-
 ## Inputs
 
 1.	name: `im_data` , shape: [1x3x480x480] - An input image in the format

models/intel/instance-segmentation-security-0083/description/instance-segmentation-security-0083.md

@@ -26,8 +26,6 @@ Feature Pyramid Networks block for feature maps refinement.
 Average Precision (AP) is defined and measured according to standard
 [MS COCO evaluation procedure](https://cocodataset.org/#detection-eval).
 
-## Performance
-
 ## Inputs
 
 1.	name: `im_data` , shape: [1x3x480x640] - An input image in the format

models/intel/instance-segmentation-security-1025/description/instance-segmentation-security-1025.md

@@ -24,8 +24,6 @@ SERes detection head, and dual attention segmentation head.
 Average Precision (AP) is defined and measured according to the standard
 [MS COCO evaluation procedure](https://cocodataset.org/#detection-eval).
 
-## Performance
-
 ## Inputs
 
 1.	Name: `im_data`, shape: [1x3x480x480] - An input image in the format

models/intel/landmarks-regression-retail-0009/description/landmarks-regression-retail-0009.md

@@ -24,8 +24,6 @@ Normed Error (NE) for i<sup>th</sup> sample has the following form:
 
 where N is the number of landmarks, _p_-hat and _p_ are, correspondingly, the prediction and ground truth vectors of k<sup>th</sup> landmark of i<sup>th</sup> sample, and d<sub>i</sub> is the interocular distance for i<sup>th</sup> sample.
 
-## Performance
-
 ## Inputs
 
 Name: "data" , shape: [1x3x48x48] - An input image in the format [BxCxHxW],

models/intel/license-plate-recognition-barrier-0001/description/license-plate-recognition-barrier-0001.md

@@ -32,8 +32,6 @@ Chinese license plates in traffic.
 Only "blue" license plates, which are common in public, were tested
 thoroughly. Other types of license plates may underperform.
 
-## Performance
-
 ## Inputs
 
 1.	name: "data" , shape: [1x3x24x94] - An input image in following format [1xCxHxW]. Expected color order is BGR.

models/intel/machine-translation-nar-en-ru-0001/description/machine-translation-nar-en-ru-0001.md

@@ -24,8 +24,6 @@ The quality metrics were calculated on the wmt19-ru-en dataset ("test" split in
 | BLEU                      |        21.6 % |
 
 
-## Performance
-
 ## Input
 
 name: tokens

models/intel/machine-translation-nar-ru-en-0001/description/machine-translation-nar-ru-en-0001.md

@@ -24,8 +24,6 @@ The quality metrics were calculated on the wmt19-ru-en dataset ("test" split in
 | BLEU                      |        22.8 % |
 
 
-## Performance
-
 ## Input
 
 name: tokens

models/intel/pedestrian-and-vehicle-detector-adas-0001/description/pedestrian-and-vehicle-detector-adas-0001.md

@@ -25,8 +25,6 @@ Average Precision (AP) metric is described in: Mark Everingham et al.
 
 Tested on challenging internal datasets with 1001 pedestrian and 12585 vehicles to detect.
 
-## Performance
-
 ## Inputs
 
 Name: `input`, shape: [1x3x384x672] - An input image in the format [BxCxHxW],